Phone a Friend: How to Get Cited by ChatGPT and Google AI Overviews

Law Smith

Founder & President, Tocobaga

Law Smith is the Founder and President of Tocobaga, a Tampa-based ROI-focused marketing agency and SMB advisory. He has strategically advised 1,000+ small and medium businesses, executed 600+ integrated marketing campaigns, and written 30+ business plans across industries ranging from professional services to e-commerce to B2B technology. Tocobaga is a Google Partner and Squarespace Gold Partner headquartered in Ybor City.

On the night of November 19, 1999, a man named John Carpenter sat across from Regis Philbin under the hot blue lights of a television studio, one question away from a million dollars, with his lifelines still unused. * Who Wants to Be a Millionaire had premiered on ABC that August as a two-week stunt hosted by a morning-show host. By the end of its first full season it would average 28.8 million viewers an episode, the top-rated show in America, pulling as many as 30 million people a night, three nights a week. * The lifelines were the show's genius: the 50:50, the audience poll, and the phone call to a smarter friend. They were meant to be spent under pressure. Carpenter had not touched one.

Then, with the million-dollar question on the screen, he asked to use Phone a Friend. The studio went quiet. The producers dialed his father. And Carpenter did not ask his dad for the answer. He told his dad he already knew it, and that he was about to win. * Then he won, the first person anywhere in the world to take the top prize on the show. *

The scene is famous because of the swagger. The reason it belongs at the top of a post about AI search is the lifeline itself. Phone a Friend was built on one assumption: that a person with 30 seconds and a question would call a human being who knew the answer. The show retired the lifeline in 2010, because by then the friends on the other end of the line were not remembering the answer. They were typing the question into a search engine while the clock ran. * The friend had become a middleman for Google.

In 2026 the middleman is gone. The friend is the machine. As of October 2025, 800 million people a week were phoning ChatGPT, by OpenAI's own count. † Many millions more get an AI Overview at the top of a Google result without asking for one. When a managing partner in Tampa asks her phone "who handles commercial lease disputes near me," or a CFO at a PE-backed HVAC platform asks "best call tracking software for a multi-location service business," the friend answers in a paragraph. It names three or four businesses. Then the buyer moves on.

This is the question Tocobaga, the lead generation agency and SMB advisory that Law Smith runs out of Ybor City in Tampa, hears more than any other right now, in more or less these words: how does a business get the friend to say its name? The honest answer is that nobody controls the friend. But the friend is not random, either. The friend has habits, and the habits have been measured. The measurements point to a short list of signals a business controls, the same list Tocobaga works through on its own site at SolvingHow.com and on client sites, and to a way of telling within a quarter whether any of it is working.

What does it mean to get cited by ChatGPT or Google AI Overviews?

Getting cited means an AI engine names a business, or links to one of its pages, inside a generated answer. It is the difference between being the friend's answer and being one of ten blue links the friend skipped on the way to the answer.

There are two doors into that answer, and most owners only know about one. The first door is the model's memory. Every large language model was trained on an enormous copy of the open web, which means a business's website, reviews, directory listings and press mentions were likely swept into the model's training data long before the owner thought about any of this. The second door is live retrieval. When a question needs current information, ChatGPT runs a web search, reads a handful of pages, and composes an answer with links. Google's AI Overviews and AI Mode do the same thing on top of Google's index. Memory decides whether the engine has heard of a firm. Retrieval decides whether it finds the firm's page today, for this question, and trusts it enough to quote.

The ancient world ran a version of this. In the third century BC, under Ptolemy II, the Library of Alexandria operated a policy that would make a modern copyright lawyer faint: any ship docking in the harbor had its books confiscated, carried to the library, and copied by official scribes. The library kept the originals. The owners sailed away with the copies. ‡ Ptolemy III went further, borrowing the official Athenian manuscripts of Aeschylus, Sophocles and Euripides against a deposit of 15 talents of silver, then keeping the manuscripts and telling Athens to keep the money. ‡ Estimates of the collection run anywhere from 40,000 scrolls to several hundred thousand, which is to say nobody knows, which is also true of training data. ‡

Here is the detail that matters for a business. The librarians at Alexandria did not just hoard. They labeled. The library's copies of Homer were tagged to record where each manuscript had come from, so a scholar pulling a scroll off the shelf knew its provenance. ‡ A scroll with no tag was just papyrus. The AI labs have already done the confiscating. Every business is already in the library. The open question is whether the scroll has a label the engine can read: a clear name, a clear location, a clear description of what the business does and for whom, repeated the same way everywhere. That is the whole game, and it is a game a two-partner law firm can win against a 200-lawyer firm, because the label is about clarity, not size.

How often do people actually see AI answers instead of a list of links?

Often enough to change behavior, and not as often as the headlines suggest. In a Pew Research Center study of 900 U.S. adults who agreed to have their browsing tracked in March 2025, 18 percent of all Google searches produced an AI summary, and 58 percent of the participants saw at least one that month. §

The behavior change is the part an owner should sit with. On result pages that carried an AI summary, 8 percent of visits ended with a click on a traditional link, versus 15 percent on pages without one. Only 1 percent of visits clicked a link inside the summary itself. And 26 percent of the time, a page with an AI summary was the last thing the person did in that session, compared with 16 percent for a plain results page. § Google disputed Pew's methodology, which is what a company does when a study lands on the wrong side of its earnings call. The direction of the finding has been replicated too many times to wave away.

How often the summary shows up depends on the question. Pew found that one- or two-word searches triggered a summary 8 percent of the time, while searches of ten words or more triggered one 53 percent of the time, and searches that began with a question word did so 60 percent of the time. § Translation: the typed-in brand name still gets ten blue links. The actual buyer question ("what does a Florida estate plan cost for a blended family") gets the friend.

Semrush's tracking of more than 10 million keywords across 2025 shows the dial moving in both directions. AI Overviews appeared on 6.49 percent of tracked queries in January 2025, climbed to 24.61 percent by July, and settled at 15.69 percent in November as Google pulled back. ‖ The mix changed more than the volume did. In January, 91.3 percent of the queries that triggered an AI Overview were informational; by October that share had fallen to 57.1 percent, while commercial-intent queries roughly doubled their share, to 18.57 percent. ‖ The summaries are migrating from "how does a 1031 exchange work" toward "who should handle a 1031 exchange." That is the migration a business should care about.

The click cost is measurable. Ahrefs compared 300,000 keywords, half with an AI Overview and half informational queries without one, across March 2024 and March 2025, and found the top-ranking page's clickthrough rate fell 34.5 percent when an AI Overview sat above it. ¶ For a regional 3PL that built a content program around "what is cross-docking" style queries, that is a third of the traffic to its best pages, gone, with the rankings unchanged. The page still ranks first. First just pays less.

On the local side the shift is steeper. BrightLocal's 2026 survey of 1,002 U.S. consumers found 45 percent had used an AI tool for a local business recommendation in the prior 12 months, up from 6 percent a year earlier. ChatGPT led at 31 percent, with Google's AI Mode at 23 percent. ** Among people aged 30 to 44, the prime hiring years for a lawyer, a CPA, a contractor or an advisor, 64 percent had asked an AI for a business recommendation in the past year. ** A jump from 6 percent to 45 percent in twelve months is not a trend line. It is a cliff, and the business is standing on it whether it looks down or not.

What do AI engines actually cite, and why is it not the same as ranking first on Google?

AI engines cite pages that answer the specific sub-question they are working on, and they draw far more from outside the top ten than most owners assume. Ranking first still helps. It is no longer sufficient, and it was never the mechanism.

The clearest evidence is a pair of Ahrefs studies on Google's own AI Overviews. In July 2025, Ahrefs found roughly 76 percent of pages cited in an AI Overview also ranked in the top ten for the same query. By March 2026, with a larger sample of 863,000 keywords and 4 million cited URLs, that overlap had fallen to 38 percent. †† Ahrefs cautions that its parsing improved between the two studies, so part of the drop is better measurement. The rest is a mechanism Google has confirmed, called query fan-out: the engine breaks one question into several related sub-questions, searches each one, and cites the pages that show up most often across those sub-searches. †† A page that ranks eleventh for the big question but first for three of the sub-questions wins. That is new, and it is winnable by a small firm with specific pages.

Which domains get cited most? Ahrefs looked at about 78.6 million searches across Google AI Overviews, ChatGPT and Perplexity in June 2025. Wikipedia was the single most-cited domain in all three, drawing 16.3 percent of ChatGPT mentions, 12.5 percent of Perplexity's and 8.4 percent of AI Overviews'. YouTube pulled 16.1 percent of Perplexity citations and 9.5 percent of AI Overviews, but did not crack ChatGPT's top ten. Reddit took 7.4 percent of AI Overview citations and appeared in neither of the other two engines' top ten. ‡‡ ChatGPT leaned toward news outlets such as Reuters and the Associated Press. AI Overviews leaned toward Reddit, Quora and health systems like Mayo Clinic. ‡‡ Pew's independent sample found Wikipedia, YouTube and Reddit together made up 15 percent of the sources in AI summaries, and that AI summaries linked to .gov sites three times as often as regular results did. §

Read those lists as a personality profile. The friend trusts encyclopedias, forums where real people argue, video, government sources and wire services. The friend trusts corroboration. It does not trust a business's homepage on its own say-so any more than a person would. For a law firm, that means the Florida Bar listing, the Avvo profile, the Google Business Profile, the local news quote and the Reddit thread where a former client says "we used them and they were straight with us" are all part of the citation, even though the firm wrote none of them.

Then there is age. Ahrefs analyzed roughly 17 million citations across ChatGPT, Perplexity, Gemini, Copilot and Google in mid-2025 and found the average page cited by an AI assistant was 1,064 days old, against 1,432 days for the average page in Google's organic results. ChatGPT's cited pages were about 458 days newer than the organic results for the same questions, roughly 15 months. §§ The friend prefers the scroll that docked recently.

Put the three findings together and the shape of the job appears. The engines cite specific answers to specific sub-questions, from sources that are corroborated elsewhere, and that have been touched recently. None of those three things is a keyword. All three are things a business can control, and the list of what Tocobaga has watched work across 600+ campaigns is exactly that list, in that order.

Does adding statistics and quotes really make AI engines more likely to cite a page?

Yes, and the effect was measured before most marketers had heard the term. In late 2023, researchers at Princeton University and the Indian Institute of Technology Delhi released the study that named the field generative engine optimization, and presented it the following August at the ACM KDD conference in Barcelona. ‖‖

The team built a benchmark of 10,000 real-world queries and ran the same source pages through nine different edits, then measured how much of the generated answer drew on each page. Three edits moved the needle more than anything else. Adding quotations from credible sources was the best performer, lifting the position-adjusted visibility score from a 19.3 baseline to 27.2. Adding statistics reached 25.2. Adding citations to sources reached 24.6. The paper summarizes the top methods as boosting visibility by up to 40 percent. ‖‖ The worst performer was the one every SEO agency sold in 2012. Keyword stuffing scored 17.7, below the page nobody had touched at all. ‖‖

Think about what that means mechanically. A generative engine is writing a paragraph. It needs raw material: a number it can repeat, a quote it can attribute, a claim it can trace. A page that says "we are a leading full-service firm" gives it nothing to hold. A page that says a Florida probate involving real property typically runs the better part of a year and that the statutory attorney fee on an estate of a given size is a specific dollar figure gives it two handles and a reason to name the firm that wrote it. The study's authors were not writing marketing advice. They were describing which pages are easiest for the machine to quote. Easiest to quote is the whole ballgame.

For an owner, the rule is simple enough to tape above the monitor. Every page that is meant to be cited carries at least one specific number, one named source for that number, and one sentence that reads like something a person actually said. The number cannot be made up, which is inconvenient, and the source cannot be the business itself, which is more inconvenient. The inconvenience is the moat. A competitor who will not do the reading cannot copy the page.

Why does entity clarity matter more than keywords for AI search?

Because an AI engine has to decide what a business is before it can decide whether to recommend it. Keywords tell a search engine what a page is about. Entities tell a generative engine who is speaking, where they are, and whether the same organization keeps showing up with the same facts attached.

Google's own documentation for AI features is almost anticlimactic on this point. It says there are "no additional requirements to appear in AI Overviews or AI Mode," that any page indexed and eligible for a snippet is eligible for the AI features, that no special structured data is required, and that whatever structured data a site does use must match the visible text on the page. ¶¶ Owners read that and relax. They should read it again. No special requirements means the engine is not waiting for a trick. It is waiting for the basics to be true and consistent, which, on most small business websites, they are not.

Entity clarity in practice means five things. First, a real About page: the founder's full name, a photograph that is not stock, the year the business started, the city it operates in, and what it does in one sentence a stranger could repeat. Second, the exact same business name, street address and phone number on the website, the Google Business Profile, the state corporate filing, the bar or licensing listing, the chamber listing and every directory, down to whether "Suite" is abbreviated. Third, basic Organization schema on the homepage and Person schema for the founder or partners, which is a few lines of JSON-LD that say in machine language what the About page says in English. Fourth, no duplicate declarations: Squarespace already generates its own Article structured data on every blog post, and pasting a second Article block into code injection hands the engine two conflicting descriptions of the same page, which is worse than none. Fifth, one author name on every post, with a bio that lists what that author has actually done.

Consider how that plays out for a PE-backed platform that has rolled up eleven HVAC companies across three states. Each acquired company still has its old name on its old Google profile, its old phone number in four directories, and a half-migrated website pointing to a parent brand nobody has searched for yet. To a generative engine, that is not one strong entity. It is eleven weak ones and a ghost. The fix is unglamorous: a single entity map, a decision about which name lives where, and a quarter of cleanup. The pattern across platform companies is always the same. Citations do not start until the engine can tell who it would be citing.

The library analogy holds here too. Alexandria's scrolls were useful because each carried a tag naming where it came from. A business whose name is spelled three ways, whose address changed two years ago in half the places it appears, and whose About page is a paragraph of adjectives, is a scroll with the tag torn off. The engine may have it. The engine cannot use it.

How should a service page be written so an AI engine can lift the answer?

A service page gets lifted when its first two sentences answer the buyer's real question in plain words, and when every section underneath can be read on its own without the rest of the page. The engine is assembling a paragraph from fragments. The page has to be built in fragments that survive being pulled apart.

Start with the heading. The old search habit was a heading like "Commercial Litigation Services." The friend does not get asked for commercial litigation services. The friend gets asked "how much does it cost to sue a contractor in Florida" or "is a lawyer needed to break a commercial lease." Headings phrased as those questions, in the words a buyer uses rather than the words the practice area committee approved, are the single cheapest change on this list. Ask Jeeves tried to teach the internet to ask whole questions back in the dial-up era, complete with a cartoon butler, and eventually retired the butler because nobody wanted to type that much. A generation later, people are typing whole questions to a machine that finally understands them. The butler was early. The format was right.

Then answer. The first two sentences under every question heading state the answer directly: the range, the timeline, the yes or no, the condition. Nuance follows. A page that opens a section with "There are many factors to consider" has told the engine, and the reader, that the answer is somewhere else. Google's guidance for AI features says structured data should match the visible text and that pages should follow the same fundamentals as any helpful page. ¶¶ The fundamentals in question are: say the thing, say it first, say it so it can be quoted.

Self-containment is the habit nobody has. Every passage should make sense if it is the only passage the engine reads. That means no "as discussed above," no "see the section on fees," and no pronoun chains that run four paragraphs without restating the subject. A passage that says "the firm" or "a Shopify store" or "Tocobaga" by name, every time, is a passage an engine can lift without breaking it.

A short format, applied to a page about estate planning for a two-partner Tampa firm, looks like this. Heading: "How much does an estate plan cost in Florida?" First sentence: a flat-fee will package at the firm runs from a specific low number to a specific high number, and a revocable trust package runs higher, with the firm's own actual prices stated. Second sentence: the price depends on whether the plan includes a trust, how many properties are involved, and whether a business interest needs to be addressed. Then the story, the exceptions, the statutory fee reference, the number from a bar association survey with its source. Then the next question. Six to ten such sections make a page that both a nervous buyer and a language model can use.

The comparison below sets the old search habits next to the signals that get a business named.

Old search habits next to the signals AI engines reward.

Signal	Old search habit	What gets a business named by AI
Headings	Practice area labels ("Commercial Litigation Services")	The buyer's actual question, in the buyer's words, with the geography attached
Opening lines	"There are many factors to consider"	The answer in the first two sentences: a number, a range, a yes or no
Evidence	Keyword density and "leading" adjectives	One statistic, one named outside source and one quotable sentence per page
Page structure	Long sections that depend on each other ("as discussed above")	Self-contained passages that make sense if lifted alone, with the business named each time
Entity	A tagline, a stock photo and three spellings of the name across the web	One name, address, phone and description everywhere; a real About page; Organization and Person schema; no duplicate Article schema
Off-site proof	Bought links and a burst of identical five-star reviews	Answered reviews, association listings, earned media, one video, honest forum presence
Index coverage	Google only; Bing never verified; AI crawlers blocked wholesale	Bing Webmaster Tools verified, IndexNow on, OAI-SearchBot allowed, Google snippet-eligible
Freshness	Publish once; change the date when traffic drops	Real quarterly revisions of the ten core pages with a true reviewed-on date
Measurement	Rankings and sessions	Ten questions, three engines, every quarter, plus an AI option on the intake source field

Do reviews and mentions on other websites help a business get named by AI?

They are, in many cases, the deciding factor. An AI engine corroborates. Before it recommends a business for a question with money on the line, it looks for the same business described the same way by sources the business does not control.

The consumer side of this is now measured. BrightLocal's 2026 survey found that among people who had used an AI tool for a local recommendation, 63 percent trusted the recommendation and only 10 percent distrusted it. ** But the same people did not stop reading reviews. A full 97 percent said they at least sometimes double-check an AI recommendation against real reviews, and 88 percent said they fact-check the AI's sources. ** Across all consumers, 42 percent said they trust an AI platform's recommendation as much as a review; among people who had actually used the tools, that rose to 64 percent. ** The AI answer and the review page are not competing. They are two halves of one verification, and the business has to be legible in both.

The engines' own source preferences tell the same story from the other side. Reddit took 7.4 percent of AI Overview citations in Ahrefs' June 2025 sample, Quora another 3.6 percent, and YouTube 9.5 percent. ‡‡ Those are places where someone other than the business describes the business. An AI Overview for "best 3PL for a Shopify brand in the Southeast" is as likely to quote a Reddit thread of ecommerce operators comparing warehouses as it is to quote any warehouse's own site.

The practical move is to put the business's actual facts into the places the friend already reads. A detailed Google Business Profile with the services list filled in and the owner answering reviews by name. A listing on the industry association's directory with the same description as the About page. A short interview on a local podcast or in a trade newsletter that states the firm's specialty in the firm's own words. A YouTube video, even an ugly one, in which a partner explains one thing well. And, for the brave, an honest presence in the Reddit or Facebook group where the firm's buyers already ask each other for recommendations, offered as an answer rather than an ad. The point is not volume. It is the same name, the same specialty and the same city, showing up where the engine goes to check.

One warning, because the earlier post on fake and removed Google reviews covered the mechanics: manufactured corroboration is detected and penalized, by Google and by the models, and a wave of identical five-star reviews reads as noise to both. Earned mentions are slower. They also survive.

Does Bing matter for showing up in ChatGPT?

Yes. A business that has never looked at Bing in its life needs to look at it now. ChatGPT's live web search does not run on Google's index.

OpenAI's documentation is careful about naming names, but not that careful. It says ChatGPT search at times works with other search providers, and it points readers to Microsoft's privacy policy to understand how those providers process a query. *** That is the closest OpenAI comes to saying Bing out loud. The same documentation lists two conditions for a site to be eligible to appear in ChatGPT search: allow OpenAI's crawler, OAI-SearchBot, to crawl the site, and make sure the host or content delivery network is not blocking traffic from OpenAI's published crawler addresses. It adds, in four words every owner should memorize, "Placement is not guaranteed." ***

The to-do list that falls out of that is short. First, claim the site in Bing Webmaster Tools, which is free, and submit the sitemap; a surprising number of small business sites have never been verified there. Second, turn on IndexNow if the website platform or CDN supports it (a number of hosting platforms do), so new and updated pages are pushed to Bing rather than waiting to be discovered. Third, check the robots.txt file and any security plugin or firewall to make sure OAI-SearchBot is not being blocked by a rule written to stop AI training crawlers; blocking GPTBot (the training crawler) and blocking OAI-SearchBot (the search crawler) are different decisions with different consequences, and a well-meaning developer in 2023 may have blocked both. Fourth, for the Google side, keep the page indexed and eligible for a snippet, because Google states that is the entire requirement. ¶¶

For a lobbying firm or a consultancy whose buyers are C-suite executives using ChatGPT inside a corporate account, this is not a nice-to-have. The CEO is not going to open Bing. The CEO's chief of staff is going to ask ChatGPT who the top three government affairs firms in Florida are, and ChatGPT is going to quietly ask Bing. A site that Bing has never properly crawled is a scroll that never docked.

How fresh does content need to be for AI engines?

Fresher than the content most businesses are running on, and the gap is measurable. AI assistants cite pages that are, on average, about a year younger than the pages Google's organic results surface for the same questions.

The numbers come from Ahrefs' analysis of roughly 17 million citations. Across ChatGPT, Perplexity, Gemini and Copilot, the average cited page had been published 1,064 days earlier, about 2.9 years. The average page in Google's organic results had been published 1,432 days earlier, about 3.9 years, which makes the organic result set 25.7 percent older. The gap on last-updated dates was smaller but ran the same direction: 909 days for AI-cited pages versus 1,047 for organic. ChatGPT's cited pages, specifically, were around 458 days newer than Google's organic results for the same questions. §§

What this does not mean is that a business should change the date on an old post and call it fresh. The engines read the content, not just the timestamp, and a page whose only change is the date is a page whose only signal is a lie. What it does mean is that the core pages, the ten or so that answer the questions buyers actually ask, need a real revision on a schedule: new numbers, a new example, a new paragraph on whatever changed in the law or the market, and a visible reviewed-on date that is true. For a manufacturer, that might be a revised lead-time section every quarter. For a law firm, a statute change the week it takes effect. For a SaaS company, a pricing page that actually reflects pricing.

The friend gets asked about now. A page that says "in 2022, the typical cost was" has already told the engine to look for a better page. The 15-month gap in ChatGPT's preferences is, in plain terms, the difference between a firm that revised its pages last spring and one that has not touched them since the pandemic.

How does a business measure whether AI engines are naming it?

By asking the engines the same questions on a schedule, and by giving every inquiry a way to say the AI sent it. No dashboard does this well yet. A spreadsheet and a calendar reminder do it fine.

The protocol Tocobaga uses is deliberately dumb. Write down ten questions a real buyer would ask, in the buyer's words, including the city or region where it matters: "who is a good construction defect lawyer in Tampa," "which 3PLs near Lakeland handle frozen food," "what should a 20-person consulting firm pay for a CRM." On the first business day of each quarter, ask each question in ChatGPT, in Google (looking for an AI Overview or using AI Mode), and in one more engine, Perplexity or Gemini, in a fresh session, logged out where possible. Record three things for each: whether the business was named, whether one of its pages was linked, and which competitors were named instead. Thirty answers a quarter. The score is the share of the thirty in which the business appears. The competitor column is the content calendar for the next quarter, because every question where a rival is named and the business is not is a page that needs to exist or needs to be rewritten.

Then close the loop on the inquiries. Every contact form, intake script and booking page gets a required source question with a specific option for "ChatGPT or another AI assistant," next to Google, referral, social and the rest. Self-reported attribution is imperfect and it is also the only attribution that survives an AI answer, because the person who read a ChatGPT paragraph and then typed the firm's name into a browser shows up in analytics as direct traffic, which is to say as nothing.

The arithmetic for a two-partner firm looks like this. The firm fields 38 inquiries a month, 114 in a quarter. The intake form shows 11 of the 114 chose the AI assistant option, 9.6 percent. Of those 11, four became clients. At the firm's average first-year client value of $2,900, that is $11,600 in a quarter from a channel that did not exist on the firm's reports a year earlier, or $46,400 annualized if the rate holds. The entity cleanup, the ten rewritten pages and the Bing setup cost the firm roughly $2,500 of outside help and perhaps eight partner hours, once, plus two hours a quarter for the measurement. Those are illustrative numbers, not a client's, and the point of them is the ratio, not the dollars: the fix is a one-time cost and the channel compounds.

The conversion side makes the ratio better than it looks. Semrush's analysis of AI-referred traffic found that the average visitor arriving from an AI search engine was 4.4 times as valuable as the average visitor from traditional organic search, measured by conversion rate. ††† The explanation is obvious once said: the friend already did the explaining. The person who clicks through from an AI answer is not researching. That person is deciding. Eleven of those are worth more than a hundred who landed on a blog post by accident.

What is the step-by-step system to get cited by ChatGPT and AI Overviews?

Eight steps, in order, most of them cheap. The order matters because the later steps do nothing until the earlier ones are true.

• Step 1, fix the entity. One legal name, one address, one phone number, one description of what the business does and for whom, identical on the website, the Google Business Profile, state filings, licensing boards and every directory. A real About page with the founder's name, photo and history. Organization schema on the homepage, Person schema for the principals, and no duplicate Article schema on a Squarespace blog.

• Step 2, pick the ten questions. Write down the ten questions buyers actually ask before hiring, in their words, with the geography attached where it matters. These are the pages. These are also the measurement.

• Step 3, build or rewrite one page per question. The heading is the question. The first two sentences answer it with a number, a range or a yes or no. Every section under it stands alone.

• Step 4, load each page with evidence. At least one specific statistic, one named outside source and one quotable sentence per page, following what the Princeton and IIT Delhi study measured: quotations, statistics and citations lifted visibility by up to 40 percent, while keyword stuffing fell below the baseline. ‖‖

• Step 5, get corroborated. A complete Google Business Profile with reviews answered by name, an association directory listing with the same description, one earned mention in local or trade media, one video explaining one thing, and an honest presence where buyers already ask for recommendations.

• Step 6, open the Bing door. Verify the site in Bing Webmaster Tools, submit the sitemap, turn on IndexNow where available, and confirm OAI-SearchBot is not blocked. Keep every core page indexed and snippet-eligible on Google. ¶¶

• Step 7, keep it fresh for real. A quarterly revision of the ten pages with new numbers and a true reviewed-on date. No date-only edits.

• Step 8, measure every quarter. Thirty answers, three engines, ten questions. A required source field on every intake form with an AI option. One spreadsheet. The competitor column becomes next quarter's content list.

An owner who does steps one through three and nothing else will usually see the first citation inside two quarters. Steps four through eight are what turn one citation into a channel. Across 1,000+ advisory conversations, the most common failure Tocobaga sees is not a business skipping the list. It is a business starting at step five, buying mentions and reviews, with an entity so muddled that the mentions have nothing to attach to.

Is AI search a replacement for SEO, a Google Business Profile and everything else?

No. It is a new distribution layer on top of the same foundation, and businesses that treat it as a replacement will spend a year building for an engine that was going to read their basics anyway. The same page that ranks on Google, converts a visitor and earns a review is the page the friend quotes.

The foundation has not changed because the engines are built on it. Google says the requirement for AI Overviews and AI Mode is that a page be indexed and eligible to show a snippet, nothing more. ¶¶ ChatGPT's search runs on a third-party index and a crawler that has to be allowed in. *** Both engines corroborate against the reviews, directories and forums that local search has rewarded for a decade. An owner who already has a complete Google Business Profile, a fast website, real photos, a working intake process and a reviews habit is most of the way there. An owner who has none of that and is now shopping for "AI optimization" has been sold a roof with no house under it.

What has changed is the size of the layer. Semrush projects that AI search traffic has the potential to overtake traditional organic search traffic within two to four years, and that for the digital marketing topics it tracks, the crossover could arrive by early 2028. ††† The timing can be argued. The direction cannot.

This is the point at which an agency conversation usually turns into a pitch, so here is the un-pitch. Most businesses can do the eight steps themselves in a quarter with a spreadsheet, a free Bing account and a willingness to write down real prices. The ones who want a second set of eyes on the entity map, the ten pages and the quarterly scorecard can call Tocobaga for an intro call at (813) 534-1261, or email hello@solvinghow.com through SolvingHow.com. The work is flat fee, never a percentage of ad spend, the contract never auto-renews, every page, account and schema line belongs to the client from day one, and nothing Tocobaga builds carries Tocobaga's name on it. It is also not for everyone. A firm that just needs its Google Business Profile finished will be told so, and pointed to the checklist for that.

Back to the studio in November 1999. John Carpenter did not need Phone a Friend. He knew the answer and used the call to tell his father so. * Almost nobody who calls the friend today is John Carpenter. They are the other kind of contestant: a person with a real problem, a short clock, and a question they cannot answer alone, who is about to trust whatever name comes back down the line. The business cannot pick up that phone. It can make sure that when the friend starts talking, it has a clear, current, corroborated reason to say the business's name. The data says the friend will. Final answer.

How does a small business get mentioned by ChatGPT?

A small business gets mentioned by ChatGPT when its identity is unambiguous across the web, its pages answer specific buyer questions in the first two sentences, and other sources corroborate it. ChatGPT's live search draws on a third-party index that OpenAI's documentation ties to Microsoft, so the site should be verified in Bing Webmaster Tools and should allow the OAI-SearchBot crawler. Pages with statistics, quotes and named sources were cited measurably more often in the Princeton and IIT Delhi study that defined generative engine optimization.

What is generative engine optimization and how is it different from SEO?

Generative engine optimization, or GEO, is the practice of making content more likely to be used and cited inside AI-generated answers from tools like ChatGPT, Perplexity and Google AI Overviews. Traditional SEO targets a ranked list of links. GEO targets a composed paragraph, so it rewards direct answers, quotable statistics, cited sources, clear entity information and freshness. The two overlap heavily: a page must still be indexed and eligible for a snippet before any AI feature can use it.

Does Bing matter for showing up in ChatGPT search results?

Yes. OpenAI's help documentation says ChatGPT search works with other search providers and refers readers to Microsoft's privacy policy, which points to Bing's index. A site that Bing has not crawled is at a disadvantage in ChatGPT regardless of how it ranks on Google. The practical fixes are verifying the site in Bing Webmaster Tools, submitting a sitemap, enabling IndexNow where the platform supports it, and confirming that OAI-SearchBot is not blocked in robots.txt or by a firewall.

How long does it take to get cited by Google AI Overviews?

There is no fixed timeline, but businesses that clean up their entity information and rewrite their core pages as direct answers commonly see a first citation within one to two quarters. Google states there are no special requirements beyond being indexed and snippet-eligible, so the delay is usually recrawling plus the time it takes for corroborating sources, such as reviews and directory listings, to align. Measuring with the same ten questions each quarter shows whether the trend is moving.

Can schema markup get a website into AI Overviews?

Not on its own. Google's documentation for AI features says no special structured data is required and that any structured data a site uses must match the visible text on the page. Schema helps an AI engine resolve the entity, which is why basic Organization schema on the homepage and Person schema for the principals are worth adding. Duplicate or conflicting schema, such as a second Article block on a Squarespace blog that already generates one, can confuse the engine rather than help it.

How can a business check whether ChatGPT or Google AI recommends it?

Ask the engines directly, on a schedule. Write ten questions a real buyer would ask, including the city where relevant, and run each one in ChatGPT, Google AI Overviews or AI Mode, and a third engine such as Perplexity at the start of every quarter. Record whether the business was named, whether a page was linked, and which competitors appeared instead. Add an "AI assistant" option to the source question on every intake form so the leads that arrive from AI answers can be counted.

Sources

*

Wikipedia. "Who Wants to Be a Millionaire (American game show)." Premiere on ABC August 16, 1999, hosted by Regis Philbin; 1999 to 2000 season average of 28,848,000 viewers; John Carpenter's win on November 19, 1999, including his Phone a Friend call to his father; the lifeline's 2010 removal because friends were using search engines. https://en.wikipedia.org/wiki/Who_Wants_to_Be_a_Millionaire_(American_game_show)

†

TechCrunch. "Sam Altman says ChatGPT has hit 800M weekly active users." October 6, 2025. Figure announced at OpenAI's DevDay keynote. https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/

‡

Wikipedia. "Library of Alexandria." Ptolemy II's decree that books on ships entering port be copied, with originals kept and copies returned; Ptolemy III and the Athenian tragedies deposited against 15 talents; collection estimates from 40,000 scrolls upward; Homeric manuscripts tagged by origin. https://en.wikipedia.org/wiki/Library_of_Alexandria

§

Pew Research Center. "Google users are less likely to click on links when an AI summary appears in the results." July 22, 2025. Analysis of 68,879 searches by 900 U.S. adults in March 2025; 8 percent versus 15 percent click rates; 26 percent versus 16 percent session-ending rates; source composition of AI summaries. 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/

‖

Semrush (Jana Garanko). "Semrush AI Overviews Study: What 2025 SEO Data Tells Us About Google's Search Shift." Updated December 15, 2025. Tracking of more than 10 million keywords; AI Overview trigger rates of 6.49 percent (January 2025), 24.61 percent (July) and 15.69 percent (November); intent mix shifts. https://www.semrush.com/blog/semrush-ai-overviews-study/

¶

Ahrefs (Ryan Law). "AI Overviews Reduce Clicks by 34.5%." April 17, 2025. Comparison of 300,000 keywords, March 2024 versus March 2025, using aggregated Google Search Console clickthrough data. https://ahrefs.com/blog/ai-overviews-reduce-clicks/

**

BrightLocal. "Local Consumer Review Survey 2026: AI and Trust." March 10, 2026. Survey of 1,002 U.S. adults; 45 percent used AI for local business recommendations, up from 6 percent; ChatGPT 31 percent, Google AI Mode 23 percent; trust and verification figures. https://www.brightlocal.com/research/lcrs-ai-trust/

††

Ahrefs (Louise Linehan). "Update: 38% of AI Overview Citations Pull From The Top 10." March 2, 2026. Analysis of 863,000 keyword SERPs and 4 million AI Overview URLs, compared with the roughly 76 percent overlap found in July 2025; query fan-out explanation. https://ahrefs.com/blog/ai-overview-citations-top-10/

‡‡

Ahrefs (Patrick Stox). "The Top 10 Most Cited Domains by ChatGPT, Perplexity, and AI Overviews in 78.6M Searches." June 11, 2025. Wikipedia, YouTube, Reddit and Quora citation shares by engine. https://ahrefs.com/blog/top-10-most-cited-domains-ai-assistants/

§§

Ahrefs (Ryan Law). "New Study: AI Assistants Prefer to Cite 'Fresher' Content (17 Million Citations Analyzed)." July 28, 2025. Average age of AI-cited pages (1,064 days) versus organic results (1,432 days); ChatGPT citations about 458 days newer. https://ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/

‖‖

Aggarwal, Pranjal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande. "GEO: Generative Engine Optimization." Proceedings of the 30th ACM SIGKDD Conference (KDD '24), Barcelona, August 2024; arXiv 2311.09735, first posted November 16, 2023. GEO-bench of 10,000 queries; quotation addition, statistics addition and citing sources as top methods; keyword stuffing below baseline. https://arxiv.org/abs/2311.09735

¶¶

Google Search Central. "AI features and your website." Google Search documentation stating no additional requirements exist for AI Overviews or AI Mode, that eligibility follows normal indexing and snippet rules, and that structured data must match visible page text. https://developers.google.com/search/docs/appearance/ai-features

***

OpenAI Help Center. "ChatGPT search." Eligibility conditions for sites (allowing OAI-SearchBot and OpenAI crawler IP ranges), use of other search providers with reference to Microsoft's privacy policy, and the statement that placement is not guaranteed. https://help.openai.com/en/articles/9237897-chatgpt-search

†††

Semrush (Rachel Handley). "We Studied the Impact of AI Search on SEO Traffic. Here's What We Learned." July 21, 2025. Average AI search visitor 4.4 times as valuable as a traditional organic visitor by conversion rate; projection that AI search traffic could overtake traditional organic within two to four years. https://www.semrush.com/blog/ai-search-seo-traffic-study/