Answer: There’s no single “submit your local AI index” - each of the top 4 engines is crawling for local businesses with their own crawler, their own web stack/index, and own (generally undocumented) location logic (so how-do-I-show-up-in-chatGPT = 4x questions, not 1x). See table below for engine-by-engine view, common robots.txt myth most local businesses carry, what’s in/out of your control, and 4x takeaways. All engine facts below are from the engine vendor own documentation or announcement (accessed 2026-11-04, source list at bottom of table).
How each engine works for local queries
| Engine | Crawlers | Local data sources | Location logic |
|---|---|---|---|
| ChatGPT | OAI-SearchBot (search index), GPTBot (model training, opt-out via robots.txt), ChatGPT-User (user-triggered fetches, robots.txt rules may not apply) | No disclosed local index. ChatGPT Search sometimes partners with other search providers (the help article’s partner privacy policies include Microsoft), plus an offline mode from OpenAI’s own indexed web; for restaurants, data is “powered by OpenTable, Resy, and Yelp” | User location approximated from IP address; OpenAI shares general location info (not the raw IP) with partner search providers; on-device location sharing is optional and off by default; ChatGPT Memory can also influence location |
| Gemini | Googlebot — the same crawler that indexes Google Search, under standard robots.txt rules | Google’s existing Search index, plus documented API grounding, plus Google Maps integration announced at I/O 2025 (May 20, 2025) | Standard Google local mechanisms: explicit city in the query, account, IP |
| Perplexity | PerplexityBot (search index), Perplexity-User (user-triggered, ignores robots.txt), Perplexity-Training (training, opt-out) | Per-query search: its own index plus live web retrieval; no disclosed local-data partnership | No documented location logic — city/region-level behavior is practitioner consensus, not in Perplexity’s docs |
| Claude | ClaudeBot (training, opt-out), Claude-Web (search/browsing), Claude-User (user-triggered fetches, not governed by robots.txt) | A server-side search tool runs when Claude decides it helps — “every response includes citations”; image search is “powered by Bing”; no disclosed local index | No documented location awareness — location details must come from the user’s prompt |
| Google AI Mode / AI Overviews (for contrast) | Googlebot | Google’s Search index, served as generated answers (RAG); no special requirements to appear | Standard Google local signals |
Sources (all links accessed 2026-11-04):
- OpenAI crawler docs: https://developers.openai.com/api/docs/bots + ChatGPT Search localized results: https://help.openai.com/en/articles/9237897
- Perplexity crawler docs: https://docs.perplexity.ai/docs/resources/perplexity-crawlers
- Anthropic crawler docs: https://support.claude.com/en/articles/8896518 + Claude search: https://support.claude.com/en/articles/10684626
- Gemini grounding docs: https://ai.google.dev/gemini-api/docs/google-search + Google I/O 2025 blog post about Gemini: https://blog.google/products-and-platforms/products/gemini/gemini-app-updates-io-2025/
- Google Search docs about AI features: https://developers.google.com/search/docs/appearance/ai-features
The table has some obvious insights:
- The data flow is different: ChatGPT, Perplexity and Claude are running their own search stack on the open web, whereas Gemini leverages Google’s index. And if you’re asking a local question, Google’s index is the deepest local data you’re gonna get (plus, now they document a connection to Maps).
- Only Google (not the others) has a local data partnership documented. “The AI uses Google Maps data” is a claim about Gemini, not the other 3.
- None of these 4 is publicly sharing the factors they’re considering for inclusion of businesses in a shortlist.
The robots.txt myth: user-triggered bots ignore your block file
Biggest mistake local businesses make (tend to be ‘23 old school SEO “block AI crawlers to prevent content theft”). Blocking AI crawlers in robots.txt with the hopes of not being included in AI answers.
This won’t work anymore. It was silently broken for you the moment you got a customer ask a question. (OpenAI is pretty clear. ChatGPT-User is for user-initiated fetching. ‘robots.txt rules might not apply’. OpenAI, accessed 2026-11-04 Same for Perplexity-User (Perplexity docs). Claude-User works the same (Claude)).
Consequences.
- If engine needs your page to answer a customer’s question, they’re gonna fetch it regardless if you allow it or not through robots.txt. By blocking OAI-SearchBot or PerplexityBot, you can potentially remove yourself from being in the engine’s prebuilt index (may be important depending on answer paths). By blocking *-User, nothing happens. Because the user-initiated path will bypass your robots.txt.
- What you do have control over is crawl eligibility. Ensuring you’re not blocked by robots.txt or serving a JS only shell to regular HTTP clients for the index crawlers (Googlebot, OAI-SearchBot, PerplexityBot, ClaudeBot). If your site serves a JS shell to bots, but serves content to browsers, what the engine sees to fetch is an empty page. See AI Crawlers and robots.txt post for full robots.txt and crawl access audit.
- What you don’t have control over is selection. Being crawlable and citable gets you into the pool of possibilities for being considered. But who gets selected (and not be one of the 3 other people that were also in the pool) is determined by…something. Who knows, all 4 engines don’t document it. (Difference of being fetchable vs recommended). And why the measurement post is important before you go spending money on ‘fixes’.
What Google tells you to do (and what it says to all engines)
Its very helpful to look at google’s official webmaster advice on getting into AI answers (AI Mode / AI Overviews) [1], as its the only statement from the engine perspective of what the bar is. And its just kinda droll.
“There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” — Google Search documentation on AI features (accessed 2026-11-04)
The same docs show two specific local relevant channels, namely Google Business Profile for local search answers and Merchant Center for shopping answers. Also Google has published a more extensive AI optimization guide whose content is “Helpful, specific, and relevant content, that can be useful for a human (or model quoting the content).
Read the guidance as it’s written, it’s Google explaining their system and there’s no similar guidance published for the other engines. The transferable bit is the pattern, where local answers use the same evidence set (profile, reviews, actual local content, 3rd party corroboration) as the old school local search has done for the past 10 years. But using a much more selective selection step. The fundementals are explained in what works in 2026 for local SEO.
What you can and can’t control
What you do control:
- crawl-worthiness – grants to the indexers in robots.txt and good content for non-browser users. 20 minute check: robots.txt for AI crawlers post
- google business profile – feeds into gemini/google AI features, most referenced local signal (engines). Categories, services, photos, review health, etc.
- actual local content on your domain – specific, self-contained, up-to-date pages that can be quoted in an answer. see what makes content AI-citable
- 3rd party surfaces – the directories/review sites/local pages the engines already reference. You can influence these, but you don’t control them.
What you don’t control (not published by any engine):
- Which factors determine which 2-4 businesses get onto a shortlist
- Which partner data sources are used as backing to any answer and how they are weighted
- Precise location signal. Whether it’s city level from IP, account, prompt - varies by engine
- Consistency run-to-run. same query, different businesses each run. No engine promises consistency.
Practical takeaways for a local brand
- Consider each engine as unique surfaces with unique crawl paths. Run your baseline query set against ChatGPT / Gemini / Perplexity / Claude individually. (This is what the local baseline test is designed to do). A solution to one engine’s fetch path may not have a solution for another engine.
- Focus on enabling crawl access first (only a yes or no lever). Unblock the index crawlers. Ensure you get content with a bare HTTP fetch. Do the audit. This part is the same across engines.
- All the fundamentals matter the same across all the engines. Things like profile data, reviews, particular local pages and corroboration support all of them. Google has said the threshold is the same content quality as they have always required. Here’s the list of stuff to do. Here’s the list of fixes in order of strength of evidence.
- Don’t trust any claims about “local data”. You’re going to hear “Foursquare is driving 70% of the local results in ChatGPT”. We couldn’t find a primary source for that number so we’re not including it in anything we publish. (You should do the same test with anything vendors claim is “inside” an engine).
There’s no guarantee you’ll get placed in an answer. But you’ll have the right map. Four engines. Four fetch paths. One set of fundamentals. A baseline for measuring. (Monthly). You can get a baseline run against a named set of competitors with tracking if you want. We document the process and the reporting in the methodology.