ChatGPT doesn't show my business. The 20-minute diagnostic.
Five checks, run in order, that find which cause is hiding you — before you spend a dollar fixing the wrong one.
Jarvis Report · 2026-08-26 · built from live public data — we never invent a score
You asked ChatGPT the question your customers ask, and it named 3 competitors while your business — real, established, better than at least one of the names on that list — appeared nowhere. The phrase owners type next is chatgpt doesn't show my business, and what they find is mostly generic advice that skips the only question that matters: which cause is hiding you? Guess wrong and you will spend months fixing the visible thing while the actual blocker sits untouched. This page is the diagnostic, not the lecture — five checks, in order, about 20 minutes.
Before you start: get a clean reading
Run every check below in a fresh chat session with no history — an assistant that remembers you answers differently than it answers your customers. And run the buyer question more than once: answers vary between runs, so one absence proves little, while absence across four or five fresh runs is a real signal. The full measurement discipline is its own guide — the AI visibility check — but fresh sessions and repeated runs are the two rules you cannot skip.
Check 1 — Refusal or miss? (2 minutes)
Look at what the assistant actually did. If it declined to name anyone — "I can't recommend specific businesses" — that is a refusal, and it says nothing about you; some question phrasings trigger caution. Rephrase as a buyer would ("who do locals use for X in Y?") and try again. If it confidently named rivals and skipped you, that is a miss — the assistant had names and yours lost. A miss means the checks below will find something. Collapsing these two states into one feeling of "AI ignores me" is the single most common self-diagnosis error we see.
Check 2 — Can bots read your website? (5 minutes)
AI crawlers generally run no JavaScript. If your site is a modern app, the crawler may receive a nearly empty shell where you see a full page. We caught our own site doing exactly this — every page serving one identical 5,034-byte empty frame to AI crawlers while humans saw the whole product. The test: view your homepage's raw source (right-click, "View page source" — not the developer tools) and look for your actual sentences in the HTML. If your services, city, and phone number are not in that raw text, machine visitors are reading a business card with nothing on it. If this check fails, stop here — this is your cause, and no listing or review work outranks it.
Check 3 — Is your identity one entity or fragments? (5 minutes)
Open your website's contact page, your Google Business Profile, and your top three directories side by side. Compare name, address, and phone character by character: "& Sons" versus "and Sons", suite numbers present or absent, a tracking number on one directory, an old address on another. Engines merge public mentions into one entity using exactly these fields, and every mismatch splits you into fragments that each look like a smaller, weaker business. Each listing looks fine alone; the damage only exists in the comparison — which is why owners never find this one without doing the comparison.
Check 4 — Does your proof lose the comparison? (4 minutes)
Take the rivals ChatGPT actually named — not the ones you consider competitors — and put their review count, rating, and most-recent-review date next to yours. Shortlists lean on legible, comparable proof, and a 4.9 with 40 reviews reads as less proven than a 4.4 with 4,000. Uncomfortable but useful: the machine is judging the paper trail, not the work. While you are there, note what their reviews mention — assistants mine review text for the "known for" phrases they attach to recommendations. Your map profile is load-bearing for this whole check; the field-by-field detail is in do AI chatbots use Google Business Profile.
Check 5 — Are you on your category's surfaces? (4 minutes)
Ask the assistant one more question: "what are the best sites to find a [your category] in [your city]?" The sources it names — booking platforms, practitioner directories, review sites, local round-ups — are the surfaces it draws on for your category. Open each and search for yourself. Absent from two or more of the surfaces the assistant itself considers authoritative is a gap that presence everywhere else does not compensate for.
Reading your results
| Result | Your cause | First move |
|---|---|---|
| Check 1: refusals only | Phrasing, not visibility | Re-test with buyer phrasing before concluding anything |
| Check 2 fails | Crawler-invisible website | Prerender, server-render, or publish plain-HTML pages — before all else |
| Check 3 fails | Fragmented identity | Pick canonical name/address/phone; correct every surface to match |
| Check 4 fails | Thin comparable proof | Start the review system today; recency compounds |
| Check 5 fails | Missing from category surfaces | Close the two or three gaps the assistant itself named |
| Several fail | Compound — common | Fix in the order above; it is sequenced by leverage |
| All five pass, still missed | Reputation gap | Earn third-party mentions — press, round-ups, community threads |
Two honest caveats. First, industries weight these causes differently — a salon's surfaces are not a probate lawyer's, and the six-causes guide has the per-industry table. Second, the manual diagnostic has a resolution limit: it tells you which cause, not how far behind you are or whether last month's fix moved anything. That takes repeated measurement on a calendar.
The 1-minute version of all five
Every check above is automatable, so we automated them: Jarvis Report runs the live buyer questions for your city across assistants — repeated runs, refusals recorded separately from misses — reads your site the way the crawlers read it, compares your identity data across surfaces, and pulls the live numbers of the rivals AI actually names. One scan, about a minute, free, from live public data. Anything unmeasured drops out of the weighting rather than being scored as zero. What stays free and what the dashboard adds: pricing.
Run the 5 checks automatically — free scan →Live public data only. We never invent a score.
Questions owners ask us
ChatGPT describes my business fine when I ask by name. Why doesn't it show me for buyer questions?
Recognition and recommendation run on different evidence. Answering "tell me about Smith Plumbing" needs only a readable description somewhere. Surviving into "who should I call for a burst pipe" means beating other businesses on comparable proof. Being known is check 2 territory; being chosen is checks 3 through 5.
Chatgpt doesn't show my business but it shows a worse competitor. What does that tell me?
Something specific and useful: your competitor's public evidence is stronger or cleaner than yours in at least one of the five areas, because that is all the assistant can weigh. Run checks 3 and 4 against that exact competitor — the gap you find is your work list, and it is usually narrower than owners fear.
I fixed my cause. When will ChatGPT show my business?
Retrieval-read surfaces — your site, listings, profiles — can reflect fixes within weeks. Reputation and training-data footprints move slower. Re-run the buyer questions monthly with identical wording and judge the trend across runs; anyone promising a specific date is guessing.
Is being absent from one assistant's answers the same as being absent from all?
No. ChatGPT, Gemini, Copilot, and Perplexity retrieve differently, and businesses are routinely visible on one and missing on another. Diagnose on at least two assistants before deciding which cause is yours — a split result usually points at surfaces (check 5) rather than fundamentals.