6 min read · updated 5 August 2026
Why isn't my practice showing up in ChatGPT?
Short answer
The most common cause is an unclaimed Bing Places listing, because ChatGPT resolves local queries largely through Bing. Other frequent causes are missing Foursquare data, review volume below roughly 30 reviews at 4.3 stars, no local editorial presence, and AI crawlers blocked in robots.txt.
Start by ruling out the boring one
Before diagnosing anything sophisticated, check whether you are blocking the crawlers.
Open yourdomain.com/robots.txt on the live domain — not the file in your repository, not what your developer says is deployed. Look for any Disallow rule affecting:
GPTBot · OAI-SearchBot · ChatGPT-User · PerplexityBot · ClaudeBot · Google-Extended · Bingbot
This matters more than it sounds. Several hosting platforms and CDNs now prepend their own managed rules to whatever you serve, so an origin file can be perfect while the live file blocks the engines. We found exactly this on our own domain the week we launched — the file we wrote was correct, and the file the internet saw was not.
If crawlers are blocked, nothing else on this list will help until that is fixed.
The six causes, ranked
1. Bing Places is unclaimed
The most common finding, and the most consequential. ChatGPT and Copilot both resolve through Bing’s index. If you have never claimed the listing, you are largely invisible to two of the four major engines regardless of how strong your Google presence is.
Check: search your practice name on Bing. Is there a complete business panel with hours, services and photos? Effort: about an hour per location. Free.
2. Foursquare data is missing or wrong
Foursquare remains a significant structured local-data provider, and almost no healthcare operator thinks about it. Stale addresses and closed-location records propagate further than people expect.
Check: search your practice at foursquare.com.
3. Review volume is below the threshold
Observed behaviour puts the practical entry point near 30 reviews at 4.3 stars or better, and above 100 in competitive metros.
Recency is weighted about as heavily as volume. Two hundred reviews with nothing in six months tends to lose to sixty updated weekly.
Check: count reviews per location on Google and Yelp, and note the date of the most recent one.
4. No third-party editorial presence
Engines treat local “best of” roundups as trusted third-party sources. In audits we run, roughly a fifth of competitor citations trace back to exactly this kind of coverage.
This is the phase most agencies skip entirely, because it is slow, manual and unglamorous.
5. No community footprint
Perplexity in particular leans heavily on community discussion — around a quarter of its citations in recent measurement. If your practice is mentioned in no relevant thread in your metro, you have effectively removed yourself from one engine’s primary source.
The obvious shortcut here is a trap. Astroturfing is detectable, it is against every platform’s rules, and when it is caught the damage substantially exceeds the benefit.
6. No structured data on location pages
Schema does not make an engine name you. It helps an engine work out which entity a page describes — which practice, which practitioner, which service, at which address. Without it, engines are inferring from unstructured text.
Check: run a location page through Google’s Rich Results Test.
Diagnose in the right order
The list above is ordered by frequency, which is also roughly the right order to work in. Claim the listings first — it is cheap, fast, and removes the most common cause. Review velocity starts next because it takes the longest to compound. Editorial and community work follow. Structured data can happen in parallel.
What you should not do first is rebuild the website. It is the most expensive intervention and, for this particular problem, usually among the least effective.
Measure before and after, or you are guessing
Fix things without a baseline and you will never know which change moved the needle — or whether anything did.
Record, per prompt: were you named, who was named instead, in what position, and which source did the engine cite. That last column is the actual diagnosis. Everything else is symptom.
Run each prompt three times. Outputs vary between runs, and a single result is close to meaningless.
What nobody can tell you
No AI platform publishes how it selects sources. Everything here is measured behaviour and documented correlation, not confirmed ranking factors. Treat anyone offering guaranteed placement in an AI answer with considerable suspicion.