Quick answer
To get your business recommended by ChatGPT, make it easy for an assistant to find, read and trust: allow its search crawler, keep your name, address, phone and services identical everywhere, publish clear service, area and price pages, and earn mentions on other sites. This is generative engine optimization (GEO), and nobody can guarantee a mention, including us.
TorontoA customer in North York, Scarborough, Etobicoke, East York, Downtown and the Beaches who asks an assistant for a recommendation gets names drawn from what it can find. We build sites for businesses across Toronto and keep each business's name, phone and hours identical on every page. We work with you by video and on site where it helps.
What does it mean to be recommended by ChatGPT?
It means an assistant names your business when a customer asks for one. There is no directory to join and no paid placement inside answers. OpenAI says placement is not guaranteed, so nobody can promise a mention, including us.
This is Part 3 of our series on AI search; Part 1, SEO vs AEO vs GEO, explains how the three fit together.
The work is called generative engine optimization (GEO): making your business likely to be named or cited when an AI assistant writes an answer. OpenAI offers no listing to claim (its Get Quote beta is for approved partners), and says its ads do not influence answers (Ads in ChatGPT).
Training data and live search: which one you can influence
An assistant answers from what its model learned in training, from a live web search, or both. A live search reads your pages as they are today, so it is the part you can influence.
OpenAI's crawler documentation separates the two: GPTBot collects pages for training, and OAI-SearchBot lets a site appear in ChatGPT search answers.
No vendor publishes how much answers vary. Judge your visibility on the same five questions asked every month.
How do AI assistants like ChatGPT find a business to recommend?
ChatGPT, Gemini, Perplexity, Claude and Copilot each pair a trained model with a live web search: ChatGPT's own search and partner providers, Google Search for some Gemini answers, Perplexity's own index, Claude's web search and Bing for Copilot.
| Assistant | Where current information comes from | Crawler or token to allow |
|---|---|---|
| ChatGPT | Web search, sometimes through partner search providers; answers may include citations | OAI-SearchBot |
| Gemini | Google Search results, for some answers | Googlebot, with Google-Extended not blocked for grounding |
| Perplexity | A real-time search for every question; numbered citations in every answer | PerplexityBot |
| Claude | A web search tool when current facts help; citations in every searched answer | Claude-SearchBot and Claude-User |
| Copilot | A query sent to Bing; a Sources section lists the queries and sites used | Bingbot |
Illustration: how an AI assistant decides what to recommend
An assistant reads the question, recalls or searches, gathers the sources it can reach, weighs them and answers with a few names and its sources.
- The question. A customer asks in their own words.
- Recall or search. It answers from training data, a live search or both.
- Gather sources. It reads the pages its search can reach; OpenAI and Anthropic say blocking their search bots hides or reduces your pages in their answers.
- Weigh what it found. The Bing Webmaster Guidelines ask for consistent names, key facts early and links from relevant websites. No AI assistant's documentation lists reviews; Google lists them for its local results, and the rest is SEO practitioners' advice.
- Answer with a few names. It may list its sources.
Which levers have real support?
Five levers have vendor support, at least in part: a crawlable site, consistent business details, clear pages on who you are and what you offer, mentions and reviews elsewhere, and fresh, direct answers.
Prices are our advice, and reviews are documented only for Google's local results. Use this checklist; the readiness score at the top of this page turns it into twelve checks.
- Let the right crawlers in and keep pages indexable. Google says AI Overviews and AI Mode need a page that is indexed and eligible for a snippet, with no special schema (AI features and your website). Microsoft asks you not to block Bingbot, whose crawling Copilot shares. The free SEO audit checks robots.txt, the sitemap and indexability.
- One name, address, phone and description everywhere. Google asks businesses to appear as they are consistently known (Guidelines for representing your business on Google), and Bing says current Bing Places details keep a business eligible for its AI answers (Bing Webmaster Blog). See local SEO basics and Google Business Profile vs website.
- Pages that say who you are, what you do, where, and what it costs. An about page, one page per service, one per area and one for prices, with key facts near the top, as the Bing Webmaster Guidelines ask. What a contractor website needs lists the pages in order.
- Mentions and reviews on other sites. Google says local prominence draws on how many websites link to you and how many reviews you have (Tips to improve your local ranking on Google). Getting onto local roundup lists for your trade is SEO practitioners' advice, not vendor documentation. Ask for reviews on more than one platform, and never pay for them.
- Fresh, direct answers. Keep prices, hours and services current, and answer each customer question directly; Part 2 of this series shows the heading-and-answer pattern.
- Brand searches, as a result to watch. No vendor documents searches for your name as an input; treat them as a sign the other levers work.
Should you block AI crawlers from your website?
Allow the bots that power search and answers (OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User, Bingbot, Googlebot). Blocking training bots (GPTBot, ClaudeBot, Google-Extended, Applebot-Extended) is a separate choice, and some vendors use one bot for both.
Allowing OAI-SearchBot is what lets ChatGPT search show your pages. GPTBot only concerns training.
AI crawlers come in three kinds: search bots, training bots and user fetchers acting for a person. OpenAI says its training and search settings are independent and that ChatGPT-User may not follow robots.txt (OpenAI).
The verified table of AI crawlers and what each one does
These are the main AI crawlers and robots.txt tokens their makers document, each linked to the company's own page.
| Agent and company | What it is for | If you block it |
|---|---|---|
| GPTBot, OpenAI | Training | Content should not be used for training |
| OAI-SearchBot, OpenAI | ChatGPT search | Not shown in ChatGPT search answers, apart from navigational links |
| ChatGPT-User, OpenAI | Visits a page for a user's request | Robots.txt rules may not apply |
| ClaudeBot, Anthropic | Training | Future material left out of training |
| Claude-SearchBot, Anthropic | Claude's search results | Not indexed for search; may reduce visibility |
| Claude-User, Anthropic | Fetches pages when a person asks Claude | Not retrieved; may reduce visibility |
| PerplexityBot, Perplexity | Perplexity's search index; not for foundation models | Page text not indexed; the domain, headline and a brief summary may remain (Perplexity) |
| Perplexity-User, Perplexity | Fetches a page during a user's question | Generally ignores robots.txt |
| Google-Extended, Google | A token with no crawler of its own: Gemini training and grounding | No effect on Google Search; Gemini grounding can stop |
| Applebot-Extended, Apple | A token: whether pages Applebot crawls can train Apple's models | Pages stay in Apple's search results; no effect on ranking |
| Bingbot, Microsoft | Bing's crawler, whose index Copilot uses | Microsoft asks you not to block it (Bing Webmaster Guidelines) |
| Amazonbot, Amazon | Improves Amazon products and services; may train Amazon AI models | Honoured within about 24 hours |
| CCBot, Common Crawl | An open archive of web pages anyone can analyse | Disallowed pages are not fetched |
| Meta-ExternalAgent, Meta | Training AI models and indexing for Meta products | Blocks both uses; changes take up to 24 hours |
We found no documentation from ByteDance for the crawler called Bytespider, so it is not in the table.
On buildsites.io we allow every crawler in this table, training bots included, and block only Bytespider, because it is undocumented. We want any assistant that might recommend us to be able to read our pages, and we will not let in a crawler whose purpose and rules nobody publishes.
When one crawl serves both search and training
Blocking a training bot can also cost you answers where one control covers two jobs, as Google-Extended does for Gemini.
Meta and Amazon answer through separate agents (Meta-WebIndexer, Amzn-SearchBot), so blocking Meta-ExternalAgent or Amazonbot leaves those alone (Meta, Amazon).
Google's crawler documentation says Google-Extended does not affect whether a site is included in Google Search and is not a Google Search ranking signal. AI Overviews and AI Mode follow Googlebot's rules and the snippet controls (our reading of Google's pages), and Search Console's Search generative AI control can exclude a site from both; Part 2 covers them.
OpenAI may use one crawl for both of its bots when a site allows both (OpenAI). Anthropic says it is rolling out a "no AI training" signal from an internet standards draft (IETF), and that robots.txt rules naming ClaudeBot take priority over it (Anthropic).
The robots.txt helper writes the lines for any of these choices.
Is llms.txt worth publishing?
llms.txt is a Markdown file that lists your key pages for AI tools. Google says it neither harms nor helps in Google Search. OpenAI, Anthropic and Perplexity do not document reading it. It is cheap and unproven; we publish one.
What it is and who proposed it
Jeremy Howard of Answer.AI proposed llms.txt on 3 September 2024 (llmstxt.org). Its site calls it a proposal: a Markdown file at /llms.txt with your name, a short summary and lists of links to key pages.
A short file for a hypothetical store, in that format:
# Example Flooring (hypothetical)
> Hardwood, vinyl plank and tile installation in Toronto and the GTA.
## Key pages
- [Hardwood installation](https://flooring.example/hardwood): price range and process
- [Service area](https://flooring.example/areas): Toronto, North York, Etobicoke
What Google, OpenAI, Anthropic and Perplexity document about reading it
Only Google has a written position: keeping an llms.txt "will neither harm nor help" your visibility in Google Search (Google). The other three do not mention reading yours.
SE Ranking, an SEO tool vendor, found the file on 10.13% of nearly 300,000 domains in a study published in November 2025, and no link between having it and how often a domain was cited in AI answers (SE Ranking).
In a study published in June 2026, Ahrefs looked at 137,210 domains that use its analytics: 28% had the file, and 97% of those files got no requests at all in May 2026 (Ahrefs).
What we do and why
We publish one at buildsites.io/llms.txt, generated from our page data, so it costs nothing to keep current; treat yours as a cheap extra, never a priority over the levers above.
The llms.txt generator builds yours from your business details.
How do you measure whether AI assistants send you customers?
In Google Analytics 4 (GA4), use the AI Assistant channel Google added on 13 May 2026, plus a custom channel whose regex catches Perplexity, Meta AI and Duck.ai. Then ask five questions in four assistants monthly, signed out.
GA4: the AI Assistant channel and a custom regex channel
GA4 files visits from recognised assistants under an AI Assistant channel with the medium ai-assistant, and counts Google's own AI Overviews and AI Mode as Organic Search.
Google's Default channel group page gives ChatGPT, Gemini, Deepseek, Copilot and Grok as examples, with no full list. Its custom channel groups page gives the steps: in a custom channel group, add a channel whose Source matches regex, paste the regex from the prompt pack below, save, and move the channel above Referral. Custom groups also apply to past data.
No assistant documents whether a citation click passes a referrer, so some AI visits may arrive as Direct.
The monthly prompt check, step by step
Ask the same five customer questions in ChatGPT, Perplexity, Gemini and Copilot once a month, signed out, in a private window, and write down who is named and which sources are cited.
- Write five questions a customer would ask, with your trade and city.
- Ask each one in all four assistants, signed out, in a private window, so your own history does not shape the answer.
- Record the date and what each answer said.
- Repeat monthly for three months, then quarterly, and judge the trend.
Our recommendation: use Search Console, GA4 and this monthly check before you pay for any AI visibility tracker.
What Search Console and Bing Webmaster Tools show about AI traffic
Search Console's Generative AI performance report counts how often links to your site appear in AI Overviews and AI Mode (Google); the clicks stay inside its normal Web totals (Google).
Part 2 covers both. Bing's AI Performance report counts how often Copilot cites your pages, not clicks (Bing).
The prompt pack turns your business type and city into five prompts and the GA4 regex.
Worked example: a Toronto flooring store (hypothetical)
A hypothetical Toronto flooring store runs its five prompts and is never named, then fixes what the check turned up in order: crawler access, consistent business details, product and area pages with prices, outside mentions and reviews. It re-checks monthly.
Two of its questions, in the customer's words:
Who installs hardwood floors in North York?
How much does vinyl plank installation cost in Toronto?
In the first check, the answers name other stores and cite lists that leave it out, its Bing Places listing shows an old phone number, and its site has no prices.
A 30-day fix list for the example store
One lever a week, starting with the two problems that can hide the store from every assistant at once: blocked crawlers and conflicting business details.
- Week 1: access and details. Search bots allowed in robots.txt; the same name, address, phone and hours on the site, the Google Business Profile, Bing Places and Yelp.
- Week 2: pages with prices. One page per product with a price range per square foot and one per area, the structure of a flooring company website.
- Week 3: three outside mentions. A supplier's installer list, a trade association and one neighbourhood roundup.
- Week 4: a review routine. A text after every job with a direct review link, on two platforms, with no incentives.
- Every month after: the prompt check. Same prompts, same assistants.
The plan does not make any assistant name the store, on any timeline.
Start with the readiness score, or book a call to go through the technical items with us.
Sources and the dates we checked them
Every product claim on this page, tools included, comes from one of these pages, listed with its own date where it prints one.
- Searching the web with ChatGPT, OpenAI (page dated 2026-10-20). Checked 2026-11-04.
- Overview of OpenAI Crawlers, OpenAI. Checked 2026-11-04.
- Ads in ChatGPT, OpenAI. Checked 2026-11-04.
- Local services Get Quote conversion spec, OpenAI. Checked 2026-11-04.
- Does Anthropic crawl data from the web, and how can site owners block the crawler, Anthropic (page dated 2026-11-04). Checked 2026-11-04.
- Enable and use web search, Anthropic. Checked 2026-11-04.
- Perplexity Crawlers, Perplexity (page dated 2026-10-29). Checked 2026-11-04.
- How does Perplexity follow robots.txt, Perplexity (page dated 2026-09-14). Checked 2026-11-04.
- How does Perplexity work, Perplexity (page dated 2026-10-26). Checked 2026-11-04.
- Gemini Apps Privacy Hub, Google (page dated 2026-10-26). Checked 2026-11-04.
- Data, privacy, and security for web search in Microsoft Copilot and Microsoft Copilot Chat, Microsoft (page dated 2026-10-29). Checked 2026-11-04.
- Which Crawlers Does Bing Use, Microsoft Bing. Checked 2026-11-04.
- Bing Webmaster Guidelines, Microsoft Bing. Checked 2026-11-04.
- Introducing AI Performance in Bing Webmaster Tools Public Preview, Bing Webmaster Blog (page dated 2026-02-10). Checked 2026-11-04.
- List of Google's common crawlers, Google (page dated 2026-10-30). Checked 2026-11-04.
- AI features and your website, Google (page dated 2025-12-10). Checked 2026-11-04.
- Optimizing your website for generative AI features on Google Search, Google (page dated 2026-07-10). Checked 2026-11-04.
- Tips to improve your local ranking on Google, Google. Checked 2026-11-04.
- Guidelines for representing your business on Google, Google. Checked 2026-11-04.
- Default channel group, Google Analytics Help. Checked 2026-11-04.
- Custom channel groups, Google Analytics Help. Checked 2026-11-04.
- What's new in Google Analytics, Google Analytics Help (entry of May 13, 2026). Checked 2026-11-04.
- About Applebot, Apple (page dated 2026-09-04). Checked 2026-11-04.
- Amazon Web Search crawlers, Amazon (page dated 2026-10-26). Checked 2026-11-04.
- CCBot, Common Crawl. Checked 2026-11-04.
- Meta Web Crawlers, Meta (page dated 2026-05-21). Checked 2026-11-04.
- The /llms.txt file, v2, Jeremy Howard, Answer.AI (published 2024-09-03, updated 2026-08-10). Checked 2026-11-04.
- Does LLMs.txt impact your AI visibility and citations, SE Ranking (published 2025-11-07, updated 2026-10-15). Checked 2026-11-04.
- We Analyzed 137K Sites: 97% of llms.txt Files Never Get Read, Ahrefs (published 2026-06-15, updated 2026-08-27). Checked 2026-11-04.
Questions and answers
Can I pay ChatGPT to recommend my business?
No. You can buy a ChatGPT ad, but not a place in the answer. OpenAI says ads do not influence ChatGPT's answers, run on separate systems from the chat model, and appear as labelled placements below a response. Nobody can buy or guarantee a recommendation.
Is there a ChatGPT business listing I can claim?
No. OpenAI does not offer a listing or directory that a business claims. Its one business-facing programme for local services, a Get Quote button, is in beta with approved partners that supply feeds of business records. ChatGPT reads your own website, if OAI-SearchBot is allowed, and the sites that already list you.
Does ChatGPT use my Google Business Profile?
OpenAI does not document Google Business Profile as a source for ChatGPT. Its help page says ChatGPT search sometimes sends queries to partner search providers and links Microsoft's and Shopify's privacy policies. Keep the profile complete anyway: Google documents it as part of its own local results.
If I block GPTBot, will my site disappear from ChatGPT?
Not from ChatGPT search. OpenAI says GPTBot is for training and OAI-SearchBot is the crawler that lets your pages appear in ChatGPT search answers, and the two settings are independent. Blocking OAI-SearchBot is what keeps you out, apart from navigational links.
How long does it take to get recommended by ChatGPT?
Nobody can promise a timeline, including us. The only documented timings are for robots.txt changes, which OpenAI says take about 24 hours to reach ChatGPT search. After the fixes, run the same prompt check monthly for three months and judge the trend.
Next step
Talk to us about your business
A 20-minute call. We'll tell you what we'd build, how long it takes and what it costs. Toronto, the GTA and Montréal.