To get ChatGPT to recommend your business, it has to be able to get into your website, read it, find you in its search sources, trust your facts, and see your page as the best answer to the customer’s question. This playbook walks through all five gates, with the exact settings and checks — the same ones we run on our own site.
Before ChatGPT names a business, it passes your company through five gates: it has to be able to get into your website, read it, find you in its search sources, trust your facts, and decide your page is the best answer. Fail any one gate and you’re out — no matter how good your work is.
When someone types “best ceramic coating near me” into ChatGPT, the assistant doesn’t “remember” a favorite shop. It runs a quick research job: it figures out where the person is, searches its sources for candidates, opens a few of their pages, cross-checks the facts and writes a short answer with links. OpenAI describes ChatGPT search as using the person’s approximate location and partnering with search providers (OpenAI Help Center).
Most guides jump straight to gate 5 — “write great content.” But in our experience the most common failure is gate 1. In our spot check of about 80 local business websites, roughly 70% didn’t let ChatGPT and Claude in at all. That’s where we start.
Only if both your robots.txt file and your firewall let the right bots in. Each AI company runs separate bots for search, for user requests and for training. Allow the search bots, decide separately about training, and check that your hosting or Cloudflare isn’t silently blocking them.
This is the single most misunderstood point. Owners hear “AI is scraping my site” and block everything — and with it, they block themselves out of AI answers. The companies document their bots separately:
| Company | Search & answers — allow | When a user asks | Training only — your choice |
|---|---|---|---|
| OpenAI (ChatGPT) | OAI-SearchBot | ChatGPT-User | GPTBot |
| Anthropic (Claude) | Claude-SearchBot | Claude-User | ClaudeBot |
| Perplexity | PerplexityBot | Perplexity-User | — |
| Google (Search, AI Overviews, AI Mode) | Googlebot | — | Google-Extended |
| Microsoft (Bing, Copilot) | bingbot | — | — |
| Apple (Siri, Spotlight) | Applebot | — | Applebot-Extended |
Sources: OpenAI; Anthropic; Perplexity; Google; Apple.
What the companies themselves say:
So the rule is simple: always allow the search bots. Whether you also allow the training bots is a business decision — blocking them doesn’t hide you from AI answers. We allow both on our own site.
robots.txt is a small text file at yoursite.com/robots.txt. A safe starting point for a local business:
# Everyone may read the site
User-agent: *
Allow: /
# AI search and answer bots — keep these allowed
User-agent: OAI-SearchBot
User-agent: ChatGPT-User
User-agent: Claude-SearchBot
User-agent: Claude-User
User-agent: PerplexityBot
User-agent: Perplexity-User
Allow: /
# Optional: opt out of AI model training only
# User-agent: GPTBot
# User-agent: ClaudeBot
# User-agent: Google-Extended
# User-agent: Applebot-Extended
# Disallow: /
Sitemap: https://yoursite.com/sitemap.xmlWatch for the classic mistake: a leftover User-agent: * + Disallow: / from when the site was in development. It tells every robot, including Google, to stay out.
Here’s where most sites actually fail: robots.txt says “welcome,” but the firewall in front of the site says “no.” Cloudflare, hosting security plugins and “block bad bots” settings often treat AI bots as threats. Cloudflare reorganized its controls in 2026 into separate Search, Training and Agent categories, and from September 15, 2026, Training and Agent bots are blocked by default on pages that carry ads. On Cloudflare, mixed-purpose crawlers that both search and train are treated by the strictest rule (The Decoder).
Note the “user” bots — ChatGPT-User and Perplexity-User — fetch a page because a real person asked, so they may not follow robots.txt at all. A firewall, however, can still block them. Perplexity explicitly advises sites with a web application firewall to allowlist its bots (Perplexity).
yoursite.com/robots.txt and look for Disallow: / next to the search bots above.Only if your key information is in the page’s HTML as real text. Most AI crawlers don’t run JavaScript, so content that appears only after scripts load, text inside images and menus in PDFs are often invisible to them.
Vercel analyzed AI crawler traffic and found that OpenAI’s crawlers, Anthropic’s ClaudeBot and PerplexityBot download JavaScript files but don’t execute them; Google’s and Apple’s crawlers do render pages (Vercel). In practice: if you open your site, press “View page source,” and can’t find your services, prices and phone number in that text, ChatGPT probably can’t either.
Only if you’re in the places AI pulls candidates from: Google’s index, Bing’s index, and the big business directories. Get indexed in both Google and Bing, submit sitemaps, and claim your listings everywhere — not only on Google Maps.
AI assistants rarely crawl the whole internet live. They start from search indexes and business databases. Microsoft announced Bing as the default search engine for ChatGPT back in 2023 (CNBC), and OpenAI today says only that ChatGPT search “partners with other search providers” — so Bing is worth treating as a likely route, not a guaranteed one. Industry analyses also point to Foursquare, Yelp and Apple Maps as major sources for ChatGPT’s local results (SurfaceLocal), and Perplexity pulls listings and reviews directly from Yelp (Maginative).
AI trusts what it can confirm in several places. Keep your name, address, phone and hours identical everywhere, mark them up with schema.org on your site, and keep a steady flow of recent reviews.
A language model is cautious about recommending something it might get wrong. If your website says you close at 7, Yelp says 6 and Apple Maps shows an old address, the safe move for the AI is to recommend someone else. Consistency is the cheapest trust signal you can build.
Structured data is a block of code that states your facts in a machine-readable form. Google’s LocalBusiness guidelines require a name and address and recommend hours, phone, URL, price range and exact coordinates; they also ask you to use the most specific business type, such as DaySpa or AutoRepair (Google). A minimal example:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "AutoRepair",
"name": "Sunset Auto Studio",
"url": "https://sunsetautostudio.com",
"telephone": "+1-408-555-0142",
"priceRange": "$$",
"address": {
"@type": "PostalAddress",
"streetAddress": "123 Example Ave",
"addressLocality": "San Jose",
"addressRegion": "CA",
"postalCode": "95112",
"addressCountry": "US"
},
"geo": { "@type": "GeoCoordinates", "latitude": 37.33874, "longitude": -121.88528 },
"openingHoursSpecification": [{
"@type": "OpeningHoursSpecification",
"dayOfWeek": ["Monday","Tuesday","Wednesday","Thursday","Friday"],
"opens": "09:00", "closes": "18:00"
}],
"sameAs": ["https://www.yelp.com/biz/...", "https://www.instagram.com/..."]
}
</script>Two honest caveats. First, Google says there’s no special schema needed for AI Overviews or AI Mode, and that structured data must match what’s visible on the page (Google). Schema isn’t a magic switch; it’s a clear label on facts that must already be on the page. Second, invented reviews or ratings in your markup break Google’s rules — and the FTC’s.
Reviews are evidence. BrightLocal’s 2026 survey found 74% of consumers only care about reviews from the last three months, and 47% won’t use a business with fewer than 20 reviews (BrightLocal). AI summaries draw on the same reviews. Add an About page with real names and photos, and get mentioned on local sites — chambers of commerce, local press, industry associations. Each independent mention is one more source that confirms you exist.
AI quotes pages that answer the customer’s exact question in plain language: what you do, for whom, where, how much, how long, and why you. Lead with the answer, then the details.
Think about how people actually ask: “Who does PPF in Campbell and how much does a full front cost?” A page titled “Welcome to Excellence” doesn’t answer that. A page titled “Paint protection film in Campbell, CA — prices and timing” with a clear first paragraph does. This answer-first format is called AEO — answer engine optimization. This very article is written that way: every section starts with a short answer.
You can’t pay for a spot inside ChatGPT’s answer, there’s no secret AI schema, llms.txt isn’t a ranking lever, and blocking training bots doesn’t hide you. The fundamentals above are what move the needle.
Ask the assistants the same questions every month, watch AI referrals in your analytics, and check Bing’s AI Performance report. No tool can guarantee placement, so measure the trend.
Week 1: open the doors. Week 2: get found. Week 3: make your facts consistent. Week 4: rewrite your key pages as answers. Then repeat monthly.
| When | What to do | Gate |
|---|---|---|
| Week 1 | Fix robots.txt, check your firewall or Cloudflare, ask ChatGPT whether it can open your site, check that key facts are in the page source | 1–2 |
| Week 2 | Google Search Console and Bing Webmaster Tools, sitemap, IndexNow | 3 |
| Week 3 | Claim and align Google Business Profile, Bing Places, Apple Business Connect, Yelp, Foursquare; add LocalBusiness markup | 3–4 |
| Week 4 | Rewrite your homepage and service pages answer-first, add prices, service area and FAQs; start a review routine | 4–5 |
| Monthly | Prompt list, analytics referrals, fresh reviews, one new answer page | all |
We don’t give advice we haven’t taken. This checklist is the one we run on brandworksapp.us and on every website we build: AI search bots allowed in robots.txt, static HTML that bots can read, structured data, answer-first pages like this one. Every site we build in our website modernization package is checked against all five gates before launch.
Want to know which gate your site fails? Our free website check tests gates 1 and 2 in about a minute. For the bigger picture of why this matters now, read Customers now ask AI where to go. If you’d rather we take it from here, that’s our AI search optimization service, and for the Maps side, local SEO and Google Business Profile.
Make sure ChatGPT can reach and read your website (allow OAI-SearchBot and check your firewall), get indexed in Google and Bing, keep your name, address, phone and hours identical across Google, Bing Places, Apple Maps, Yelp and Foursquare, add LocalBusiness structured data, collect recent reviews, and write service pages that answer customers’ questions with real prices and your service area. Placement is never guaranteed.
At minimum OAI-SearchBot (ChatGPT), Claude-SearchBot (Claude), PerplexityBot (Perplexity), Googlebot and bingbot, plus the user-request fetchers ChatGPT-User, Claude-User and Perplexity-User. Training bots such as GPTBot, ClaudeBot and Google-Extended are optional and don’t control search visibility.
No. According to OpenAI, GPTBot only collects training data. Whether you appear in ChatGPT search answers depends on OAI-SearchBot.
Not directly, according to industry analyses. ChatGPT relies on sources such as Bing, Foursquare, Yelp and Apple Maps, plus business websites, so keep every listing accurate — not only your Google Business Profile.
It doesn’t hurt, but don’t expect results from it. Google has said no AI system currently uses llms.txt, and Google says no special AI files are needed for its AI features.
OpenAI says robots.txt changes can take about 24 hours to reflect in ChatGPT search. Listings, reviews and content take weeks to months, and no one can guarantee placement. Track it monthly with the same set of questions.
We checked about 80 local business websites: roughly 70% block ChatGPT and Claude. How to check yours in 2 minutes.