Five playbooks, one operating system. Each book covers one system in full and then applies it to all 14 industries NetWebMedia serves, so you read the method once and then only the section written for your business.
© 2026 NetWebMedia. All rights reserved.
Published by NetWebMedia · www.netwebmedia.com · WhatsApp +1 (442) 385-4585 · English and Spanish editions
Every statistic in this book carries a numbered reference to a named, dated source in the Sources section. Benchmarks are medians or averages from other businesses as published by those sources — not predictions of your results.
This ebook is licensed to the purchaser for use within one business. Please do not share, resell or redistribute it; team and agency licences are available on WhatsApp. Thank you for supporting independent work.
The Answer Engine Optimization playbook for local and service businesses — 5 phases, 14 industries.
This playbook is for the person who owns or runs the business: the dentist who also signs the invoices, the hotel manager, the restaurant owner, the plumber with no agency, the salon owner with three chairs. It assumes a website you can edit, a Google Business Profile, four hours a month, and no marketing background.
The situation it addresses: a customer used to type something into Google, get a page of links, and your job was to be one of them. Increasingly they ask an assistant — ChatGPT, Perplexity, Google's AI Overviews and AI Mode, Gemini, Claude — and get back a short answer naming two or three businesses. If you are not named, you were not in the running. Forty-five percent of consumers have now used AI tools for local business recommendations, up from 6% a year earlier, making AI the third most-used discovery tool behind Google and Facebook and ahead of Yelp and TripAdvisor.1
By the end you will know three things you almost certainly do not know today: whether AI engines mention your business, which sources they pull from when they answer questions in your category and city, and what to publish so your name is in the answer. You will have a 30-question audit to re-run monthly, a five-phase method, a 90-day plan and a set of prompts that do most of the writing.
How to use it: read Part 1 in order — the method is the same for every business. Then read only your own industry section in Part 2; the other thirteen are not for you. Then work through Part 3, which holds the templates you will reuse: the audit, the fact sheet, the monitoring log, the page checklist.
What it does not do. It does not make you appear in AI answers on demand, because nobody controls that, and any vendor who says otherwise is selling something. The engines change behavior without notice — one cut its Reddit citations by roughly 80% inside a few days in 2025. Every benchmark here is a median or a study finding from a named source with a date, not a prediction about your business. The method removes the reasons an engine currently has to skip you, and makes you the easiest business in your category to quote correctly.
Start with the traffic you can no longer count on. In the first four months of 2026, 68.01% of US Google searches ended without a click, up from 60.45% in 2024, on SparkToro's analysis of Similarweb clickstream data.2 In the same panel, AI Overviews appeared on more than 20% of searches, and where they appeared click-through fell by nearly 60%.2 Pew Research Center measured the same effect on real browsing: users clicked a traditional result on 8% of visits with an AI summary versus 15% without, and clicked a link inside the summary on 1%.3
The audience is not small. Google's AI Mode passed 1 billion monthly active users and the Gemini app reached 950 million, per Alphabet's Q2 2026 earnings call.4 ChatGPT is reported to be approaching 1 billion weekly active users, and passed 1 billion monthly app users in May 2026 with 50 million-plus paying subscribers — reported from company data rather than an OpenAI statement, so treat it as directional.5
The number that matters: 45% of consumers have used AI tools for local business recommendations, up from 6% the year before. Among AI users, 63% trust the recommendations they get. Among 30–44 year olds, 64% have asked AI for a local recommendation. ChatGPT is used by 31%, Google AI Mode by 23%.1
That is adoption. Behavior is more nuanced, and the nuance favors a careful operator. In a separate BrightLocal survey, 23% of consumers used an AI tool during their most recent local search but only 8% started there; 18% would contact a business on an AI recommendation alone, while 43% verified it on Google first.6 AI is rarely the whole path. It is the shortlist — it decides which three names get verified.
Volume is modest and intent is unusually high. Adobe Analytics, across more than a trillion visits to US retail sites, found AI-referred traffic up 138% year over year and up 1,324% since October 2024, converting 54% better than non-AI traffic with 53% more time on site and 23% more pages per visit.7 A year earlier that traffic converted at about half the rate of everything else. People arriving from an AI answer have already been told you are a fit.
The growth sits in the queries that precede a purchase. Across more than 600,000 keywords between November 2025 and April 2026, AI Overview coverage of commercial-intent keywords grew 71% while transactional-intent coverage fell 5%.8 Commercial intent means "best dentist for implants in Santiago", not "buy toothpaste".
Two caveats. There is no single number for how often AI Overviews appear: Semrush's 10-million-keyword panel recorded 6.49% of queries in January 2025, 24.61% in July 2025 and 15.69% by November 2025, while curated commercial keyword sets report far higher.9 The gap is sampling, not contradiction. And local intent suppresses AI Overviews — one analysis of 116,918 search results pages in May 2026 found local-intent queries triggering them 24.7% of the time versus 40.1% for general queries, on Australian data, so treat it as directional.10
So: AI answers are not yet where most of your customers come from. They are where a growing minority of them get their shortlist, and where the majority will get it soon. The work is cheap, it is work you should have done anyway, and it compounds.
Watch out: appearing in Google's Local Pack does not mean appearing in Google's AI Mode. Across 1,120 query executions in 13 US cities in May 2026, 28.5% of Local Pack businesses were absent from AI Mode for the identical query; overlap was 71.5% overall but only 48% among the top three.11 Being first on the map is not evidence that you are visible to AI.
There is no ranking algorithm to reverse-engineer here, and anyone who claims to have one is guessing. What exists is measurement of where citations physically come from, plus a little official documentation.
Ranking helps. Ranking is not enough. Ahrefs found in July 2025 that 76.10% of AI Overview citations came from pages ranking in Google's top 10, across 1.9 million citations.12 Their March 2026 re-run of 863,000 keywords put that at 38%, with 31.2% of cited pages ranking 11–100 and 31.0% not ranking at all, attributing part of the swing to better parsing and Google's query fan-out.13 BrightEdge, tracking nine industries from May 2024 to September 2025, found the overlap moving the other way — 32.3% up to 54.5% — with wide sector differences: healthcare 75.3%, education 72.6%, e-commerce 22.9%, restaurants 19.2%.14 Two credible datasets pointing opposite ways means holding the range, not a number: rankings correlate with citation but do not produce it, and in low-overlap categories most citations come from outside the top ten.
Brand mentions beat backlinks. Across 75,000 brands, Ahrefs measured the correlation between AI Overview visibility and each signal: branded web mentions 0.664, branded anchors 0.527, branded search volume 0.392, Domain Rating 0.326, backlinks last at 0.218. Twenty-six percent of brands had zero AI Overview mentions, and the top web-mention quartile averaged 169 mentions against 14 for the next.15 Being named, in text, on other people's sites is the strongest correlate anyone has measured — for a local business that means chamber pages, supplier pages, sponsorships, local press, association directories and "best of" round-ups.
Most citations come from sources you already control. Yext analyzed 6.8 million AI citations across 1.6 million queries per model in mid-2025: 86% came from sources the brand controls — first-party websites 44%, listings 42%, reviews and social 8%, forums 2%. Gemini leaned hardest on websites (52.1%), OpenAI on listings (48.7%).16
That sits oddly beside the better-known finding that across 230,000+ prompts and 100 million+ citations the most-cited domains overall are Reddit, Wikipedia and YouTube.17 Both are true, and reconciling them is the whole strategy. Open-ended informational questions pull from the big community and encyclopedia sites; local, branded and "who should I hire" questions pull overwhelmingly from listings and the business's own site. You compete in the second kind. One analysis of 30 million sources also found Google's AI leaning on Facebook and Yelp, and Perplexity on Reddit.18
There are review floors. Uberall's May 2026 analysis reports the assistants applying effective rating thresholds before recommending a business: ChatGPT prioritizes 4.3 stars and above, Perplexity 4.1+, Gemini 3.9+.19 Below those lines, no amount of page-level work fixes your visibility. Fix the rating first.
Schema is hygiene, not a lever. Ahrefs tested 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 matched controls: AI Overviews −4.6%, AI Mode +2.4% and ChatGPT +2.2%, neither of the last two significant.20 Schema correlates with citation in observational studies because authoritative publishers use it. Add it — it keeps your facts machine-readable — but do not pay for it as an AI visibility service.
Google says the same in its own documentation: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary," and "You don't need to create new machine readable files, AI text files, or markup to appear in these features."21 On the file everybody is selling, John Mueller of Google stated in June 2025: "FWIW no AI system currently uses llms.txt."22 Publishing one costs nothing. Promising results from it is not defensible.
The one technical step with an official causal claim behind it is in OpenAI's crawler documentation. OAI-SearchBot surfaces sites inside ChatGPT search, and OpenAI states plainly: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links."23 GPTBot is training-only and ChatGPT-User is user-initiated; neither determines search appearance. Plenty of sites blocked all three in 2024 and never revisited it.
Do this now: open
yoursite.com/robots.txt. A block onOAI-SearchBot, or a blanketDisallow: /underUser-agent: *, means you have opted out of ChatGPT search results. Decide deliberately — blockingGPTBotto stay out of training data is legitimate and does not require blockingOAI-SearchBot.
Finally, content shape. The Princeton and IIT-Delhi "Generative Engine Optimization" paper is the best causal evidence available, tested on 10,000 queries across 25 domains. Adding relevant quotations lifted a page's position-adjusted visibility by 41%; adding statistics by 33%; citing sources by 28%; improving fluency by 27%. Keyword stuffing reduced visibility by 8–10%. Validated live on Perplexity, quotation gave +22% and statistics +37%.24 The highest-leverage edit to any page is not a keyword. It is a quotable sentence and a number you can stand behind.
Watch out: "add statistics" means publish your own verifiable numbers — response time, price range, production volume, placement rate. It is not permission to invent industry statistics. An engine that quotes a number you made up has made you the publisher of that claim, and in several industries in Part 2 that is a regulatory problem, not an embarrassment.
Five phases, run in order the first time, then looped monthly. Each has a job, a set of actions, and a check.
Write 30 questions a real customer would ask before buying from a business like yours: ten category questions with your city, ten constraint questions (price, hours, insurance, language, parking, availability), five comparisons and five branded. Part 3 has the template and ten examples to adapt.
Ask all 30 in five places: ChatGPT, Perplexity, Google AI Mode, Gemini and Claude. Log four things per question: were you named, who was named instead, which sources were cited, and whether anything said about you was wrong. Use a signed-out session — you want the generic answer.
Two numbers come out. Your citation rate is the share of the 150 answers (30 questions × 5 engines) naming your business. Your error rate is the share containing a wrong fact — wrong hours, wrong price, a service you dropped. Both go in the monitoring log.
The number that matters: your starting citation rate. Most local businesses find it in single digits, and in several categories the floor is worse: 84% of individual wedding vendors have effectively zero AI citation share,25 and 87% of independent plumbing and HVAC contractors do too.26 A very low first score is normal. It is also a baseline you can move.
Do this now: before Phase 2, list the five facts customers ask you for most on the phone. Those are the five facts your website is probably missing.
An engine can only quote what it can read as text. Three actions.
First, publish the facts you currently hide: prices or ranges, hours including holidays, service area boundaries, what is and is not included, minimums, lead times, languages, parking, accessibility, cancellation terms. The proof — across 120 gyms in 50 North American markets, AI assistants scored 48/100 on understanding those businesses, with pricing at 0.10 out of 1.0, and hedged 64% of answers with lines like "prices aren't listed, you'd want to check with them".27 A hedged answer is a lost customer. Publishing a price range is a citation strategy.
Second, get those facts out of PDFs, images and JavaScript widgets and into HTML text. A menu as a JPG, a price list as a PDF, a timetable inside a booking iframe — all invisible. Tables and short labeled paragraphs work, as do plain sentences.
Third, add Organization, LocalBusiness and FAQPage schema. Treat it as hygiene — the controlled test found no significant citation lift.20 It keeps your name, address, phone and hours consistent across surfaces, which is worth doing anyway.
Check it worked: paste your main service page into an AI assistant and ask it to list every fact it can extract. If it cannot state your price range, hours and service area from the page alone, the page is not finished.
Phase 1 gave you the questions where somebody else got named. Each is a page: one genuinely useful page per question — a reference, not a blog.
Build each page to the shape the evidence supports: a direct answer in the first two sentences, your own numbers, a quotable line, sources cited where you rely on someone else's data, a visible last-updated date. That is the GEO paper applied literally — quotations +41%, statistics +33%, citing sources +28%, keyword stuffing harmful.24 Your own data is what competitors cannot copy: real price ranges for the ten most common jobs, average wait time, pass rate, occupancy by month.
Write for a person who will never see your website. If an engine strips the page to one sentence, which sentence should it be? Write that one first and put it near the top.
Eighty-six percent of citations came from brand-controlled sources — 44% first-party websites and 42% listings.16 So the second half of the work is off your site.
Claim and complete every listing that matters in your category — Part 2 names them industry by industry — with facts identical to your site. Keep your rating above the floors in the previous chapter.19 Get onto the "best of" lists that already rank for your category question; in several industries a directory profile beats anything on your own domain. Earn plain-text mentions of your name on local sites: sponsorships, supplier pages, association member lists, local media. Google's AI Mode has been observed drawing on Business Profiles, review platforms including Yelp, and "unstructured citations" such as community pages.28
For Latin America, one extra step. Assistants collapse Spanish-speaking markets into a single generic answer, partly because so little AI investment and crawling reaches the region and because Spanish costs roughly 59% more tokens than English for the same content.29 Make market context explicit instead of writing neutral Spanish: name city and country together, use local currency, use comuna or barrio names, set the regional language tag. And remember the model here — the AI answers, the click lands on your page, WhatsApp closes the sale. Put the WhatsApp button above the fold.
Re-run the same 30 questions on the same five engines on the same day each month. Change no wording; comparability is the point. Log citation rate, engines covered, sources cited, factual errors, and what you shipped. When a number moves you will have candidate causes instead of a shrug.
If you would rather see a first pass before committing four hours, the free AI visibility checker at netwebmedia.com will run a version of the Phase 1 audit for you.
One quarter, one phase at a time, roughly four hours a week. Adjust the pace, not the order.
| Weeks | Phase | Do this | Done when |
|---|---|---|---|
| 1–2 | Entity audit | Write the 30 questions, run them on 5 engines, log every answer and source | Baseline citation and error rate recorded |
| 3–4 | Structured facts | Fix robots.txt; publish prices, hours, service area and inclusions as HTML text; add schema | An AI assistant can extract every key fact from your service page |
| 5–8 | Citation corpus | Publish 4–6 pages, one per question where a competitor was named, each with own data, a quotable line and a date | Pages live, dated and linked from the service page |
| 9–11 | Distribution | Claim and align every listing; fix the rating; pitch 3 "best of" lists; earn 5 local name mentions | Listings match your site fact-for-fact; rating above 4.3 |
| 12 | Monitoring | Re-run all 30 questions; complete the log; pick next quarter's four gaps | Month-over-month comparison in one sheet |
Watch out: do not compress all five phases into week one. The audit is only useful as a before-and-after, so record the baseline before you publish anything.
Do this now: block the same two hours on the first Monday of every month for the rest of the year and title it "AI audit". The method fails for one reason above all others: the audit gets run once.
Use your AI assistant (Claude, ChatGPT or Gemini) for the drafting. Replace everything in [brackets], and never publish what comes back without checking every fact against your own records.
Watch out: an assistant will invent a plausible price, a certification or a statistic to fill a gap you left. Every prompt below tells it not to. Check the output against your records before any of it goes live.
Do this now: paste the first prompt below into your assistant and run its questions on one engine today. Twenty minutes gives you the first rows of your log.
Generate the audit questions:
You are helping a [industry] business in [city, country] audit its visibility in AI answers.
Write 30 questions a prospective customer would type into an AI assistant before choosing a business like this:
- 10 category + city questions ("best X in Y")
- 10 constraint questions on price, hours, insurance/payment, language, parking, availability, urgency
- 5 comparison questions
- 5 branded questions about [business name]
Use the customer's words, not industry jargon. Output a numbered list, nothing else.
Turn a page into extractable facts:
Here is the text of my [service] page: [paste page text].
Act as an AI search engine. List every concrete, quotable fact you can extract: prices, hours, service area, duration, what is included, credentials, capacity, policies.
Then list the facts a customer would expect that are MISSING or vague.
Then rewrite the page so every fact appears as plain text — short labeled paragraphs or a table — with a one-sentence direct answer at the top. Invent nothing I have not given you.
Draft a FAQ from real customer questions:
These are questions my customers actually ask, from emails, WhatsApp and phone notes: [paste 15-25 real questions].
Group them into themes, then write an FAQ for a [industry] business in [city].
Each answer: 40-70 words, direct answer first, specific numbers where I gave them, no marketing language, no claims I have not supplied.
Flag any question that needs a price or range I have not given you.
Write a quotable, original-data paragraph:
Here is real operating data from my [industry] business: [paste your numbers — average job cost by type, response time, seasonal occupancy, pass rates].
Write one 60-90 word paragraph publishing this data, plus one standalone sentence an AI assistant would be likely to quote verbatim.
Be precise about what the numbers do and do not cover, state the period, and add a "last updated [month year]" line. Use no adjectives.
Audit robots.txt:
Here is my robots.txt: [paste file].
Tell me line by line which AI and search crawlers are blocked and which are allowed — specifically OAI-SearchBot, GPTBot, ChatGPT-User, Google-Extended, PerplexityBot and ClaudeBot.
Explain in plain English what each block prevents. Then give me a corrected file that keeps my site eligible to appear in AI search answers, and say which lines you changed and why.
Rank is the wrong scoreboard: a page can rank third and never be quoted, and be quoted while ranking nowhere — 31.0% of cited pages in one 2026 sample did not rank in the top 100.13 Track four things instead.
Citation rate. Named answers divided by total answers, from your 30-question audit. Watch the trend across months, not the absolute number — one question set is a sample.
Engine coverage. How many of the five engines name you at all. Being strong in one and invisible in another is common; ChatGPT leaned on listings while Gemini leaned on websites in the Yext data, so a gap points at which asset is weak.16
The number that matters: engines covered. Being named by one of the five is a start; being named by four is a position a competitor has to work to take from you.
Source mix. Each month, tally which sources the engines cited on your category questions. That tally is your work list: if a directory appears in eleven of thirty answers and you have no profile there, that is next month's job.
Assisted conversions. Watch referrals from AI domains in your analytics, and add "did an AI assistant suggest us?" to your booking form. Expect low volume and high quality: AI-referred traffic converted 54% better than non-AI traffic in Adobe's retail dataset.7 One software company reported its AI search visitors converting at 23 times its organic rate, with AI search at 0.5% of traffic but 12.1% of signups — one company's first-party data, an illustration of the shape and useless as a benchmark.30
What "working" looks like: a citation rate climbing month over month, error rate falling to zero, and your business named in the constraint questions, not just the branded ones. Being named when someone asks for you by name is table stakes; being named for "a dentist near me open Saturday who takes my insurance" is the win.
Watch out: every benchmark here is a median or average from a named study, on a specific sample and date. None is a forecast for your business, and the engines change behavior without notice. Use them to judge whether you are in the normal range, never as a promised target.
Read your own section only. Each follows the same shape: where the money leaks, five moves, the compliance caution, a first-30-days checklist. Numbers are specific to your category where a study exists; where none exists this book says so.
Guests ask assistants what your booking engine cannot answer: "best boutique hotel in Valparaíso for a couple under $150 a night", "which hotels near Napa have free parking and let dogs stay". The answer comes mostly from the OTAs. Across 19,579 model runs on six engines covering 31,138 hotels in 25 cities, TripAdvisor was cited in 99.9% of Grok responses and 95.5% of Perplexity's, Booking.com in 76.4% of Grok's.31 But direct hotel links made up 75–91% of the actual recommendations.31 The OTA is the source; your site is the destination — if it answers the question.
Do this now: ask three engines "hotels in [your city] that allow dogs" and see whether your pet policy appears. If it lives only in a PDF, it does not exist.
Accessibility claims are a legal representation in most jurisdictions, not marketing copy. If you publish "wheelchair accessible" for an AI to repeat, state what is and is not accessible: door widths, roll-in shower, lift access, step-free route from the car park. An engine quotes the general claim and drops the qualifier, so make the qualifier part of the claim.
Restaurants are the worst-served category in the dataset. Uberall found 83% of restaurant locations entirely invisible in AI recommendations, while the top three brands in a category take 53.4% of share of voice and category leaders get ten times the AI mentions of an average competitor.19 Demand is there: 22% of US diners have used AI to choose a restaurant, rising to 61% among those aged 25–34.32 Listings and review platforms account for over 41% of the sources AI cites for restaurants, with ChatGPT favoring third-party directories, Gemini restaurant websites and Perplexity reviews and forums.32 In fine dining, one index reports Eater overtaking the Michelin Guide as the dominant anchor — modeled, not measured, so take the direction only.33
The number that matters: 83% of restaurant locations are invisible in AI recommendations.19 The bar is low, and it is mostly a text-formatting problem.
Allergen and "gluten-free" claims carry food-safety liability, and in the US "gluten-free" is an FDA-defined term with a specific threshold. Never let an AI-optimized menu page overstate what your kitchen can do. If you cannot control cross-contact, publish "we can modify but we are not a gluten-free kitchen" — an engine will quote it.
Patients are moving faster than any other category. Forty-seven percent now use AI to find providers, up from 31% at the end of 2025; 36% say AI influenced their choice and 60% trust AI summaries without verifying them.34 Two-thirds hit incorrect provider information — wrong address, phone, insurance or hours.34 That last figure is the opportunity and the risk in one line. Healthcare has the tightest coupling to organic rankings of any vertical, at 75.3% of AI Overview citations from ranking pages,14 and a hard floor: 75% of patients will not consider a provider rated below 4.0 stars.34
Watch out: 60% of patients trust AI summaries without verifying them.34 If an engine repeats an old address or a plan you dropped, most people never find out. Re-run your branded questions monthly and correct the source, not the answer.
The FTC requires "competent and reliable scientific evidence" for health claims, and you cannot use a patient testimonial to imply a result you could not substantiate directly.36 That covers anything an engine may quote, including outcome and before-and-after copy. And HIPAA: never publish a patient story, photo or review response with treatment details without written authorization.
There is no beauty-specific AI citation study yet, and this book will not invent one. What we have is the cross-category number — 45% of consumers have used AI for local business recommendations1 — plus two demand-side findings: 55% of salon clients are comfortable with AI receptionists, and 69% have abandoned a booking because the business was hard to reach.37 The questions here are all constraint questions: "best salon near me for curly hair that takes walk-ins", "balayage on dark hair under $200", "barber near me open Sunday". Each needs a price, a specialism or an opening hour that most salon sites keep in an Instagram bio.
Do this now: ask an engine "how much does balayage cost at [your salon]" and see whether it hedges. If the answer is "prices aren't listed", publish the range this week.
Injectables and most aesthetic treatments are medical procedures in nearly every jurisdiction. Advertising them triggers the health-claim standard — "competent and reliable scientific evidence" — plus practitioner licensing rules.36 Name the practitioner and their registration on the page, and publish no outcome claim you cannot evidence.
If you do not fit one of the other thirteen, this is your section, and the headline finding is the general one: 86% of AI citations come from sources the brand already controls — first-party websites 44%, listings 42%, reviews and social 8%, forums 2%.16 Most of that is inside your control, and mostly undone: one study of 400 companies found 46% of owners invisible to AI, with traffic up 2.5% year over year where both owner and business were visible against −28% where neither was.38 It is agency-authored and correlational, but the invisibility rate matches every other dataset here.
The number that matters: 44% of citations come from first-party websites and 42% from listings.16 Between those two you control 86% of the surface area.
The FTC's Consumer Reviews and Testimonials Rule took effect on 21 October 2024 and bans buying, selling or creating fake reviews, conditioning compensation on a review's sentiment, undisclosed insider reviews, and legal threats to suppress negative reviews.39 It applies to every industry in this book. Given the review floors the engines apply, the temptation is real and the penalty is worse than a low rating.
Seven directories effectively own AI citations for legal queries — Chambers, Legal 500, Super Lawyers, Best Lawyers, Martindale, Avvo and Justia — with firm sites, including large national firms, appearing below directory listings even for their own practice areas.40 That research is PR-firm authored, so treat it as directional; the directory list matches how those sites rank organically. Coverage is real: 23.6% of legal queries trigger AI Overviews and 57.9% of question-style legal queries generate an AI answer.40 And the most useful tactical finding in this book is a legal one: for "divorce lawyer", bare queries returned zero AI Mode listings, but adding "near me" restored results in 38 of 38 cases.11 That study also found AI Mode favoring educational content over business listings.
Do this now: run your five core practice-area questions with and without "near me" on Google AI Mode. The difference shows which pages need explicit city and state language in the body text.
ABA Model Rule 7.1 bans false or misleading communications about a lawyer's services, including misleading omissions.41 State bars add rules on testimonials, past results and "specialist" claims. AI-generated marketing copy is the firm's responsibility, and a past-result number published for an engine to quote is a communication in the sense the rule means.
The portals hold the ground. Zillow, Realtor.com, Redfin, Trulia and Homes.com collectively account for an estimated 61% of real-estate URLs in publicly available LLM training datasets.42 Agents are almost entirely absent: only 8.4% of US agents appear in AI responses for high-intent searches in their own market, and the top 1% capture 47% of all AI citation share, across 8.2 million queries in 192 metros.42 Real estate also showed the largest local-intent AI Overview gap of any category, at −37.5 points.10 Meanwhile Zillow's share of agent-discovery traffic fell from 41.2% to 33.8%42 — the portal is losing grip faster than agents are filling the space.
The number that matters: 91% of agents are absent from AI answers in their own market.42 Showing up at all is a competitive position here.
The Fair Housing Act prohibits discrimination on race, color, religion, sex, national origin, familial status and disability, including in published statements and advertisements.44 Neighborhood guides are the classic trap. Describe schools, transit, amenities, noise and price data. Never describe who lives there, and never let an AI assistant draft "family-friendly" or "safe neighborhood" copy.
This category has the clearest published proof of what fixes citation. Across 120 gyms in 50 North American markets on four assistants, citations broke down as the business's own website 42%, other individual sites 39%, search results 5.9%, directories and booking platforms 5.3%, social 4.8%, Reddit 3%.27 Then the diagnostic: the assistants scored 48 out of 100 on understanding those businesses — services 0.73 out of 1.0, availability 0.56, pricing 0.10 — and hedged 64% of answers.27 Pricing at 0.10 is not a ranking problem, it is a publishing problem. On the demand side, AI-referred retail traffic grew 138% year over year and converted 54% better than non-AI traffic.7
The number that matters: AI assistants scored 0.10 out of 1.0 on gym pricing, and hedged 64% of answers.27 Publishing your prices is the highest-return single edit in this category.
Auto-renewing memberships are regulated — in the US by FTC negative-option rules and state auto-renewal statutes, in Chile by SERNAC consumer-protection rules. Publish cancellation terms accurately and in full: you are now publishing them for an AI to repeat to a prospect who will hold you to that version.
The research sites answer the question, then hand over the customer. Sixty-six percent of AI users say they trust car-shopping and review sites such as Cars.com most for objective vehicle information, and 41% visit the cited dealer or manufacturer website next.45 The citation is the lead. Forty-four percent of consumers opt to use AI-powered car search tools and 97% of AI users say it will affect their purchase decision, though 63% worry about bias.45 Expect Cars.com, Edmunds, KBB, CarGurus and manufacturer sites in the citations — and your own site in the click that follows.
Do this now: ask three engines "how much should a brake job cost in [your city]". If a national average stands in for your market, publish your own range this week.
US dealers are covered by the FTC's used-car Buyers Guide rule and by advertising-substantiation rules. "Certified", "inspected" and warranty claims must be exact and must match the paperwork. A price or warranty term published loosely for an AI to repeat becomes an advertised claim you must honor.
Parents have already switched. Fifty-seven percent have used AI tools such as ChatGPT to evaluate colleges, 34% said AI surfaced schools they had not previously considered, and 84% search online on their child's behalf.46 That 34% is the argument: for schools, AI is a discovery channel, not just a research tool. Education is also one of the most AI-Overview-saturated verticals, with prevalence rising from 18% to 83%,47 and 72.6% of its AI Overview citations come from organically ranking pages.14 Expect Niche.com, GreatSchools, U.S. News, Wikipedia and Reddit alongside institutions' own sites.
The number that matters: 34% of parents said AI surfaced schools they had not previously considered.46 That is enrollment volume that never touched your marketing.
US career and vocational programs face FTC and Department of Education substantiation rules on placement and earnings claims. Publishing an unverified "95% job placement" for an AI to quote is exactly the exposure — the engine repeats the number without your footnote. Publish only figures your records evidence, and state cohort, period and definition.
Two platforms take the category. The Knot (13.0%), Zola (9.5%) and WeddingWire (7.0%) together appear in roughly 73% of wedding-planning AI responses — and one company owns two of the three — across 65+ prompts on four engines in Q1 2026.25 Against that, approximately 84% of individual wedding vendors — photographers, florists, planners, venues, caterers — have effectively zero AI citation share.25 The research is PR-firm authored, so read the shares as directional. The implication is unusual and worth acting on: here, directory profile completeness beats almost any work on your own site.
Watch out: if your directory profiles are thin, website work will not rescue you here. Fix the profiles first, then build the site pages they link to.
Venue pricing is a contractual representation. Publish service charges, gratuity policy, overtime rates and cancellation terms exactly as your contract states them, because an engine will quote the published version to a couple who arrive expecting it. Where alcohol is served, license terms restrict how you may advertise it — see the wine section.
About one in five Americans who sought financial advice in the past year turned to AI — roughly 25% of Gen Z and millennials, 16% of Gen X, 7% of boomers — but only three in ten have confidence in AI for money management and just 3% trust it "a great deal", against eight in ten with some confidence in human advisers.48 That gap is the position: AI builds the shortlist, the human closes. Coverage is uneven because so many finance queries are price lookups — overall finance AI Overview coverage rose only from 6% to 21%, while educational queries jumped from 16% to 67%.49 Educational content is the door. Expect NAPFA, the CFP Board, BrokerCheck, Investopedia and NerdWallet among the sources.
The number that matters: roughly one in five people seeking financial advice used AI, but only 3% trust it a great deal.48 You are competing for a list a cautious human will then verify.
In the US, the SEC Marketing Rule and FINRA rules govern testimonials, endorsements and performance claims for registered advisers and broker-dealers, and testimonials require specific disclosures. Never publish performance figures without the required disclaimers in the same paragraph — an engine lifts the number and leaves the footnote behind.
Citation share for HVAC and plumbing queries concentrates on software platforms and national franchises, not local contractors: ServiceTitan 14.0%, Roto-Rooter 11.5%, ARS/Rescue Rooter 5.5%, Mr. Rooter 4.8%, Housecall Pro 4.2%.26 Roughly 87% of independent contractors have essentially zero AI citation visibility, and the sharpest line in that report is that "an independent plumber with 800+ five-star Google reviews in a single metro produces less AI citation share than a franchise location with 50 reviews".26 It is PR-firm research, so hold it as directional. Reviews still matter locally: among plumbers, businesses appearing only in AI Mode had 337% more reviews at the median.11
Do this now: publish the emergency triage page this week. One page, an hour's work, and it targets the highest-volume question in your category.
Trade licensing rules in most US states, and in Chile, require the license number to appear in advertising. "Licensed and insured" with no verifiable number is a common violation, and now an AI problem too: an engine cannot verify a claim with no number. Publish the number, the issuing body and the expiry.
No published AI-citation study exists for this vertical, and this book will not manufacture one. Two transferable findings apply. First, 86% of citations come from brand-controlled sources — 44% first-party websites, 42% listings — so your technical sheets and your Vivino, Wine-Searcher and Google Maps profiles are the whole surface.16 Second, Food and Drink was the fastest-growing AI Overview category in Semrush's panel, up 7.25 percentage points since March 2025.9 For direct-to-consumer sales the retail evidence applies: AI-referred traffic grew 138% year over year and converted 54% better.7 Expect Vivino, Wine-Searcher, Wine Enthusiast, Decanter and Google Maps in your citations.
Do this now: move one wine's technical sheet out of its PDF into HTML text, then ask an engine for the alcohol level and blend. That test tells you whether the rest of the catalog is worth converting.
US wine advertising under 27 CFR Part 4 Subpart G requires the responsible advertiser's name and location plus the class or type statement, and heavily restricts any statement related to health — including implying a physical or psychological sensation from drinking the wine. Health claims require scientific evidence, risk disclosure and equally prominent disclaimers.50 Chile applies its own Ley 19.925 restrictions. Never let an AI assistant draft tasting copy that strays into wellbeing claims.
Build the list once, then never change the wording — comparability across months is the point. Ten category questions, ten constraint, five comparisons, five branded. Run each on ChatGPT, Perplexity, Google AI Mode, Gemini and Claude in a signed-out session, so you get the generic answer.
Ten examples to adapt:
Log four fields per answer: named (yes/no), who was named instead, sources cited, any factual error about you.
One page on your site and one internal document, carrying the same facts — the raw material engines use to work out who you are. Fill every field, publish it as text, and use identical wording in every listing.
| Field group | Fields to publish |
|---|---|
| Identity | Legal name, trading name, founding year, owner name, tax or registration number |
| Contact | Address, phone, WhatsApp, email, booking URL, hours including holidays |
| Offer | Services with price or range, duration, what is and is not included, minimums |
| Coverage | Service area by suburb, comuna or postcode; languages; payment methods |
| Credibility | Licenses with numbers and issuers, insurance, certifications, memberships, years trading |
One row per question per engine, one sheet per month. Ten minutes to fill, and the only record that will tell you what worked.
| Column | What goes in it |
|---|---|
| Date / Engine / Question | The fixed question set, run the same day each month |
| Named | Yes or no — your citation rate numerator |
| Competitors named | Every business the answer named instead of you |
| Sources cited | Every domain shown; tally monthly to build your work list |
| Errors | Any wrong fact about you, and which source it came from |
| Shipped this month | Pages, listings and mentions added, so a change has candidate causes |
Run this on every page you publish for AI visibility.
Analyze your audit results:
Here are my AI visibility audit results for a [industry] business in [city]: [paste your log — question, engine, named yes/no, competitors named, sources cited].
Tell me: (1) citation rate overall and by engine, (2) the five domains cited most often, (3) which question types I lose most, (4) the three highest-leverage fixes in order, with a reason for each.
Be blunt about what the data does not show.
Build your entity fact sheet:
Act as a fact-extraction engine. Here is everything I know about my business: [paste raw notes, licenses, hours, prices, service area].
Organize it into a publishable fact sheet with these groups: identity, contact, offer, coverage, credibility.
Mark every field I have not given you as MISSING and fill no gap with an assumption. Then write the page as short labeled paragraphs of plain text.
Write a listing-consistency check:
Here is my website's key information: [paste]. Here is what my [Google Business Profile / directory] listing says: [paste].
List every discrepancy, however small — name, hours, phone, services, prices, service area.
Rank them by how likely each is to make an AI assistant hedge or answer wrongly, and say which source to correct.
Draft the answer page for a lost question:
An AI assistant was asked "[question from my audit]" and named [competitors] instead of my business.
Here are my real facts on this topic: [paste].
Write the page that would deserve to be cited: direct answer in the first two sentences, my own numbers, one quotable standalone sentence, a short FAQ of the follow-up questions, and a last-updated line.
No adjectives, no claims beyond the facts I gave you, and flag anything a customer needs that I have not supplied.
Run the audit once and you have a baseline. Run it monthly for a year and you have something rarer: a record of what actually moved your visibility, in your city and category. If you would rather start with an outside read, the $49 AI Visibility Scan at netwebmedia.com returns a written report within 48 hours, credited in full against AEO Starter.
BrightLocal, Local Consumer Review Survey — AI trust, 2026 — https://www.brightlocal.com/research/lcrs-ai-trust/
SparkToro / Similarweb, zero-click study, 2026 — https://sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/
Pew Research Center, 2025 — https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
Google / Alphabet Q2 2026 earnings, 2026 — https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q2-2026/
PYMNTS, citing The Information, 2026 — https://www.pymnts.com/news/artificial-intelligence/2026/chatgpt-approaches-1-billion-weekly-active-user-milestone/
BrightLocal, consumer search behavior and channels, 2026 — https://www.brightlocal.com/research/consumer-search-behavior-channels/
Adobe Analytics via Digital Commerce 360, 2026 — https://www.digitalcommerce360.com/2026/06/17/adobe-ai-referred-traffic-to-retail-sites-doubles-in-a-year/
Semrush, AI Overviews commercial-intent study, 2026 — https://www.semrush.com/blog/ai-overviews-commercial-search-study/
Semrush, AI Overviews study, 2025 — https://www.semrush.com/blog/semrush-ai-overviews-study/
Digital Nomads HQ, State of AI Search — local, 2026 — https://digitalnomadshq.com.au/state-of-ai-search/local-search/
Local SEO Data, AI Mode vs Local Pack study, 2026 — https://www.prweb.com/releases/local-seo-data-study-finds-28-5-of-local-pack-businesses-are-invisible-in-google-ai-mode-302796340.html
Ahrefs, search rankings and AI citations, 2025 — https://ahrefs.com/blog/search-rankings-ai-citations/
Search Engine Journal, reporting Ahrefs' 2026 update — https://www.searchenginejournal.com/google-ai-overview-citations-from-top-ranking-pages-drop-sharply/568637/
BrightEdge, rank overlap after 16 months of AI Overviews, 2025 — https://www.brightedge.com/resources/weekly-ai-search-insights/rank-overlap-after-16-months-of-aio
Ahrefs, brand mentions and AI Overview visibility (75,000 brands), 2025 — https://ahrefs.com/blog/ai-overview-brand-correlation/
Yext, AI citations research (6.8M citations), 2025 — https://www.yext.com/about/news-media/ai-citations-release
Semrush, most-cited domains in AI search, 2025 — https://www.semrush.com/blog/most-cited-domains-ai/
Peec AI, via Search Engine Land, 2026 — https://searchengineland.com/ai-search-engines-cite-reddit-youtube-and-linkedin-most-study-473138
Uberall, restaurant AI visibility report, via Business Wire, 2026 — https://www.businesswire.com/news/home/20260507962493/en/83-of-Restaurants-Are-Invisible-in-AI-Search-New-Uberall-Report-Reveals-the-Discovery-Gap-Reshaping-the-Quick-Service-Restaurant-Industry
Ahrefs, schema and AI citations controlled experiment, 2026 — https://ahrefs.com/blog/schema-ai-citations/
Google Search Central, "AI features and your website" — https://developers.google.com/search/docs/appearance/ai-features
John Mueller (Google) on llms.txt, via Search Engine Roundtable, 2025 — https://www.seroundtable.com/google-ai-llms-txt-39607.html
OpenAI, crawler documentation — https://developers.openai.com/api/docs/bots
Aggarwal et al., GEO: Generative Engine Optimization (Princeton / IIT Delhi), arXiv 2311.09735 — https://arxiv.org/pdf/2311.09735
5W Public Relations, wedding AI visibility research, 2026 — https://www.prnewswire.com/news-releases/73-of-wedding-planning-ai-answers-now-route-to-just-two-platforms--and-84-of-individual-wedding-vendors-have-zero-ai-citation-share-new-5w-report-finds-302766400.html
5W Public Relations, HVAC and plumbing AI Visibility Index, 2026 — https://www.5wpr.com/ai-visibility-index/hvac-plumbing-ai-visibility-index-2026/
Courtyard, State of AI Visibility — gyms, 2026 — https://getcourtyard.ai/state-of-ai-visibility/gyms
Whitespark, guide to Google's AI Mode for local businesses, 2026 — https://whitespark.ca/guides/whitesparks-guide-to-googles-ai-mode-for-local-businesses/
Search Engine Land, "global Spanish" AI search visibility, 2026 — https://searchengineland.com/global-spanish-ai-search-visibility-472833
Ahrefs, first-party AI search traffic and conversions, 2025 — https://ahrefs.com/blog/ai-search-traffic-conversions-ahrefs/
Nicolas Sitter / hotelrank.ai, AI hotel landscape, 2026 — https://www.nicolassitter.com/research/ai-hotel-landscape-2026
Bloom Intelligence, AI restaurant discovery (compiling DoorDash, Yext and others), 2026 — https://bloomintelligence.com/blog/ai-restaurant-discovery/
5W Public Relations, restaurants citation share index, via Everything-PR, 2026 — https://everything-pr.com/restaurants-citation-share-index-2026
rater8, via TechTarget, 2026 — https://www.techtarget.com/patientengagement/news/366643954/Almost-half-of-patients-use-AI-for-online-provider-search
Forbes Chile, on Doctoralia, 2026 — https://forbes.cl/negocios/2026-07-01/una-polaca-y-una-espanola-el-tripadvisor-de-los-medicos-suma-3-millones-de-visitas-al-mes-en-chile-y-hub-clave-para-ia/
US Federal Trade Commission, Health Products Compliance Guidance — https://www.ftc.gov/business-guidance/resources/health-products-compliance-guidance
Zenoti, salon and spa booking and communication trends, 2025 — https://www.zenoti.com/thecheckin/salon-spa-booking-communication-trends
Thrive Internet Marketing Agency, via Entrepreneur, 2026 — https://www.entrepreneur.com/growing-a-business/the-one-number-every-business-owner-needs-to-know/504118
US Federal Trade Commission, Consumer Reviews and Testimonials Rule — https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers
5W Public Relations, Legal AI Visibility Index, 2026 — https://www.5wpr.com/ai-visibility-index/legal-ai-visibility-index-2026/
American Bar Association, Model Rule 7.1 — https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/rule_7_1_communication_concerning_a_lawyer_s_services/
FlyDragon, State of AI SEO in real estate, 2026 — https://www.goflydragon.com/state-of-ai-seo-in-real-estate/
Similarweb, top classifieds websites in Chile, 2026 — https://www.similarweb.com/top-websites/chile/e-commerce-and-shopping/classifieds/
US Department of Justice, Fair Housing Act — https://www.justice.gov/crt/fair-housing-act-1
Cars.com, AI car-shopping survey, 2025 — https://www.cars.com/articles/cars-com-survey-reveals-ais-growing-influence-on-car-shopping-97-of-ai-users-say-it-will-impact-purchase-decisions-and-almost-half-have-already-leveraged-the-tech-for-car-shopping-518967/
EAB, via The EDU Ledger, 2026 — https://www.theeduledger.com/ai/article/15831466/more-than-half-of-parents-now-use-ai-to-research-colleges-survey-finds
SERPs.io, AI Overview prevalence by industry — https://serps.io/blog/ai-overview-prevalence-by-industry
Gallup, via Fortune, 2026 — https://fortune.com/2026/08/08/financial-advisers-trust-ai-usage-gap-gallup/
BrightEdge, AI Overviews insights for finance, 2025 — https://www.brightedge.com/resources/weekly-ai-search-insights/ai-overviews-healthcare-insights-for-finance
US eCFR, 27 CFR Part 4 Subpart G, wine advertising — https://www.ecfr.gov/current/title-27/chapter-I/subchapter-A/part-4/subpart-G
This playbook is general educational information for business owners and managers. It is not legal, financial, tax, medical or other professional advice, and reading it does not create a client relationship with NetWebMedia. Laws, platform policies and prices change; consult a licensed professional in your jurisdiction before acting on anything that has legal or financial consequences.
Statistics are reported as published by the sources listed in the Sources section, with their dates. Where a source is a vendor, a PR firm or a modeled estimate, the text says so. Benchmarks are medians or averages measured on other businesses; they describe a range, not a forecast for your business. NetWebMedia does not guarantee rankings, citations, leads, costs or revenue. Results vary.
Company, product and platform names are the property of their respective owners and are used for identification only. Worked examples are hypothetical unless a source is cited.
Everything in this playbook is work you can do yourself. If you would rather have it done, NetWebMedia runs it as a monthly service — in English and Spanish, for businesses in the United States and Latin America.
At netwebmedia.com: see in one minute whether AI engines currently find your business.
A written report within 48 hours: real queries across ChatGPT, Claude, Perplexity and Google AI, an AEO score and a 10-point fix list. The $49 is credited in full toward the AEO Starter plan.
AEO plus SEO and content strategy from $249 a month. Elsewhere, AEO/GEO retainers start at $1,000–2,500 a month according to The Digital Elevator's 2026 pricing guide.