Generative engine optimization (GEO) is the work of making a practice easy for AI answer engines to find, trust and cite. Google describes the mechanism plainly: its AI features split a question into several related searches, pull current pages from its search index, and link the pages that support the answer. That makes GEO an extension of SEO. A practice gets named when its site is easy to crawl, says specific and checkable things about treatments, providers and locations, and is backed by an accurate Google Business Profile and steady reviews.
What happens when a patient asks AI for a recommendation?
Say a patient types best place for lip filler near me into Google’s AI Mode. The engine has no private list of favorite practices. Google describes two steps. First comes query fan-out: the model breaks the question into several related searches and runs them at once, along the lines of lip filler cost, injector credentials nearby, and lip filler reviews. Then comes retrieval-augmented generation, which Google also calls grounding: the system pulls current pages from its search index for those searches, reads them, writes the answer, and links the pages that support it. [1]
So an AI answer is assembled from pages the engine found and read moments earlier. If your lip filler page is not among them, the answer gets written without you, and it can draw on the related searches as much as the original one.
The scale is already large. Google says AI Mode passed one billion monthly users a year after launch. [2] Among Americans who look up health information online, 65% have seen AI-generated answers at the top of search results, and 63% of adults consider AI health information somewhat or very reliable. [3] OpenAI says more than 230 million people ask ChatGPT health and wellness questions every week. [4] And when an AI summary appears in Google, people click a regular result on 8% of visits, against 15% when there is no summary. [5] The answer itself is becoming the place a patient decides who to call.
Does every AI engine find practices the same way?
No. Each engine reads the web through its own crawler or index, so showing up in one does not put you in the others. The practical difference comes down to which crawler you allow.
| Engine | Where it finds pages, and the crawler to allow |
|---|---|
| Google AI Overviews and AI Mode | Google’s own search index and Business Profiles. There is no separate AI crawler: if Googlebot can index the page, the AI features can use it. |
| ChatGPT search | Pages reached by OAI-SearchBot. GPTBot is a different crawler that collects training data, and blocking it does not remove you from ChatGPT search. |
| Perplexity | Pages reached by PerplexityBot, which Perplexity says surfaces sites in its results and is not used to train models. |
| Microsoft Copilot | Bing’s index, so the same Bingbot access and Bing Webmaster Tools that serve Bing search. |
| Claude | Claude-SearchBot for search and Claude-User when a person asks Claude to read a page. ClaudeBot collects training data. |
OpenAI states it directly: sites opted out of OAI-SearchBot are not shown in ChatGPT search answers. [6] Microsoft’s guidance points the same way as Google’s: the SEO practices that help Bing discover and index a page also make it eligible for AI-generated answers and citations. [7] If you want to know which of these gates is closed for your practice, our Not Showing Up in AI Search page walks through the diagnosis.
What does the research say gets a page cited?
The term GEO comes from a research paper by a team from Princeton University and IIT Delhi, presented at KDD 2024, a major data science conference. [8] The researchers rewrote web pages in different ways and measured how much of each page a generative engine used in its answers. Three changes did the most: citing credible sources, adding quotations, and adding statistics, each improving visibility by 30 to 40% on their main measure. Keyword stuffing, the old SEO habit, helped little or not at all, and on Perplexity it did 10% worse than doing nothing.
The most useful finding for a practice is who gained. A page ranked fifth that added citations to its sources saw its visibility rise 115%, while the top-ranked page in the same test lost ground. [8] Good, well-sourced content can be quoted ahead of a page that outranks it.
Two cautions keep this honest. The paper tested a 2023 research setup built on GPT-3.5 and the top five Google results, not today’s ChatGPT or AI Mode, so read it as direction rather than a promise. And rankings and citations have drifted apart since. In a study of 863,000 searches, Ahrefs found that 38% of the pages cited in Google’s AI Overviews also ranked in the top ten for the same search, compared with about 76% in its July 2025 study, noting that it also improved how it detects citations between the two. [9] Ahrefs connects the shift to fan-out searches, and found YouTube is now the domain AI Overviews cite most.
For a practice, the lesson is to cover the questions around a treatment as well as the treatment itself: cost, downtime, who performs it, what results look like, with specific facts and named sources. Put real expertise on video where you can.
What does Google say you can ignore?
In July 2026 Google updated its own guide to optimizing for AI search, and its message is direct: for Google, optimizing for generative AI search is still SEO. [1] It also lists popular GEO tactics that do nothing for Google Search:
- llms.txt and other AI-only files. Google Search does not use them. Other AI systems may, so keeping them is harmless.
- Special AI markup. No schema is required for AI features. Structured data still earns rich results in regular search.
- Chopping pages into small chunks or rewriting copy for machines instead of patients.
- Buying or faking mentions across the web.
What Google says does matter is original, expert content with a first-hand point of view, which it calls non-commodity content; a site that is easy to crawl and fast to load; good images and video; and an accurate Google Business Profile for local answers. It added one new requirement: a site has to be included in the Search generative AI features setting in Search Console, a control Google rolled out to every site on August 31, 2026, so it is worth confirming yours. [1]
Watch: Google Search Central’s Q2 2026 update (6 minutes)
John Mueller, a Search Advocate on Google’s Search Relations team, walks through the new Generative AI report in Search Console and Google’s guide to optimizing for AI search.
What is AI actually doing on a practice’s website?
We measure it. Every site we build on Urge Foundation keeps an AI Activity report that logs each AI crawler visit by company and purpose. On our own site, from October 8 to 19, 2026, AI crawlers made 5,384 requests. About 48% collected text to train AI models, 40% indexed pages for AI search, and 12% fetched a page because a person had just asked an assistant a question. None of that last group appears in Google Analytics, because an assistant reading a page does not run the analytics code. Not every request could be confirmed as coming from the company it named, which is its own lesson about trusting raw counts. It is a small, early sample, and we are preparing a fuller look across the Urge Foundation sites we run, which we will update as the data grows.
What comes after answers: assistants that act
The next step is AI doing things for the patient. At its 2026 developer conference, Google said people can ask it to call businesses on their behalf in categories including beauty. [2] Google Business Profile has a setting for these automated calls, which can ask about bookings, prices and availability. [10] Practices should decide on that setting deliberately and make sure the front desk is ready for those calls.
In the browser, the change has a name: WebMCP, a proposed web standard drafted in the W3C Web Machine Learning Community Group by engineers from Microsoft and Google. [11] It lets a website hand an assistant clear tools instead of leaving it to guess which buttons to press. Chrome and Microsoft Edge are testing it in origin trials that run through March 30, 2027. [12] Our site and several of the practice sites we build on Urge Foundation already run it in both browsers’ trials: an assistant can look up offices and hours, providers and treatments, and fill in a request form, and the patient presses send. Few assistants use WebMCP yet, and no search engine treats it as a ranking factor. We run it because being ready before patients start using it is the advantage.
How can a practice check its own AI visibility?
- Search Console. The Generative AI performance report shows how your pages appear in AI Overviews and AI Mode. While you are there, confirm the AI-features setting is on.
- Bing Webmaster Tools. Its AI Performance report shows when Copilot cites your pages, and which pages.
- A monthly question set. Ask ChatGPT, Google’s AI Mode, Perplexity and Copilot the same ten patient questions each month, by treatment, by provider name and by city, and note who gets named.
- Your robots.txt file. Make sure it allows Googlebot, Bingbot, OAI-SearchBot, PerplexityBot and Claude-SearchBot. Blocking a training crawler such as GPTBot is a separate choice that does not affect AI search.
- Your new patients. Ask how they found you. When the answer starts including an assistant, you are in the answer.
- Or see it for yourself on Urge Foundation. Every site we build on Urge Foundation includes the AI Activity report, so you can see which AI companies read your site, why, and which pages they read, alongside the visitors Google Analytics counts. Talk with us about moving your practice’s site onto it.
For a MedSpa-specific version of this checklist, see how aesthetic practices show up in ChatGPT answers.
How Urge Foundation and our team put this into practice
Urge Foundation, the platform every site we build runs on, starts with what Google documents: fast, crawlable pages with clear structure. On top of that it builds in:
- A canonical answer at the top of key pages: a short, plain statement an engine can lift as written.
- Complete schema for the practice, its treatments, providers and locations. AI features do not require it, but it earns rich results and tells every engine what the practice is.
- llms.txt, llms-full.txt and a plain-text copy of every page for the AI systems that read them. Google ignores them, and they cost nothing.
- Deliberate crawler access, with instant change notices to Bing through IndexNow so updates reach Copilot quickly.
- The AI Activity report, showing which AI companies visit, why, and which visits can be confirmed against the addresses each company publishes.
- WebMCP tools in the Chrome and Edge trials, so assistants can act on the site with the patient in control.
Our AI Search Optimization work adds the content, the Business Profile and review work, and the monthly testing, and the Urge Intelligence Platform brings Analytics, Search Console, Business Profile and ad results into one report. We update this guide as the engines change.
Generative engine optimization (GEO) is the work of making a website easy for AI answer engines, such as ChatGPT, Google’s AI Mode and AI Overviews, Perplexity and Copilot, to find, trust and cite. The term comes from a 2023 research paper by a team from Princeton University and IIT Delhi. You may also see it called AEO or AI search optimization.
Mostly not. Google says that for its own search, optimizing for generative AI is still SEO, because its AI features retrieve pages from the same index with the same ranking systems. What changes is the emphasis: AI answers draw on related searches as well as the original question, so pages that cover the full set of patient questions with specific, sourced facts tend to be cited.
Not for Google. Google says its search ignores llms.txt and that no special schema is required for its AI features, though structured data still earns rich results. Other AI systems may read llms.txt, so Urge Foundation publishes it at no cost to your Google visibility.
No. Every engine’s retrieval is its own, and Google notes that no third-party tool has access to its internal ranking or AI systems. What you can control is a site the engines can read, specific and well-sourced content, an accurate Business Profile and steady reviews, and regular testing of the questions your patients ask.
Block training crawlers if you prefer, but keep the search crawlers open. GPTBot and ClaudeBot collect training data; OAI-SearchBot, PerplexityBot and Claude-SearchBot are what put a page in AI search answers, and OpenAI says sites that block OAI-SearchBot are not shown in ChatGPT search answers.
Want to see where your practice stands in AI answers?
We’ll test real patient questions across ChatGPT, Google’s AI Mode, Perplexity and Copilot and show you where you are named, where a competitor is, and what to change.
