Checking rankings used to mean Search Console. Not anymore.
For fifteen years, "check my rankings" meant one thing: open Google Search Console, look at position and impressions, done. That workflow does not exist for AI search. ChatGPT, Perplexity and Gemini do not publish a rankings dashboard, do not show you impression data, and do not tell you which businesses they recommended to which users. There is no equivalent of Search Console for generative engines, and there will not be one anytime soon, because the whole point of a synthesized answer is that it is generated fresh for each query rather than pulled from a static ranked list.
That opacity is exactly why most local businesses have no idea whether they show up in AI search at all. They assume that if their SEO is solid, their AI visibility must be fine too. Sometimes that is true. Often it is not, because the signals that get you cited in ChatGPT (schema completeness, FAQ structure, Bing indexation, entity consistency) are different enough from classic Google ranking factors that a site can rank on page one of Google and be completely invisible in every AI engine at the same time.
The only reliable way to check is to ask the engines directly, the way a real customer would, and read what comes back. That is what this article walks you through: five exact prompts, run across three engines, with a plain framework for interpreting whatever you find.
The 5 diagnostic prompts to run
Replace the bracketed placeholders with your actual service, city, neighborhood and business name. Run each prompt exactly as written, since small wording changes can change which sources an engine pulls from.
These five are not arbitrary. Each one tests a different layer of visibility, from broad category recognition down to name-level recall, so a business can pass some and fail others in a pattern that actually tells you something. Write down the full response for each prompt, not just whether your business appeared, since the surrounding context (what the engine says about competitors, what attributes it emphasizes) is often as useful as the citation itself.
- 1. "Best [service] in [city], with pros and cons" — This is the broadest test. It tells you whether your business is even in the consideration set for a generic, high-intent search.
- 2. "Compare top [service providers] in [city]" — A comparison prompt forces the engine to name multiple specific businesses side by side, which is a better test of competitive standing than a single recommendation prompt.
- 3. "Who are the most trusted [service] companies in [neighborhood]?" — Narrowing to neighborhood level tests hyper-local visibility, which is where a lot of businesses lose ground because their content and citations are city-wide, not neighborhood-specific.
- 4. "Recommend a [service] near [address] for someone who values [attribute]" — This tests whether the engine can match your business to a specific customer need (fast turnaround, budget-friendly, licensed and insured, family-owned) rather than just a generic category match.
- 5. "What do people say about [your business name]?" — This is a direct name-recognition test. It tells you whether the engine has enough information about your specific business, independent of the category, to say anything substantive at all.
How to read the results: 4 outcomes and what each one means
Every response you get back falls into one of four buckets. Knowing which bucket you are in tells you what to fix, so do not skip this step and just eyeball the answer.
- Cited by name (win): your business is named specifically, ideally with some accurate detail about it (services, location, a specific strength). This means your technical foundations and content are working. The goal now is expanding this from one or two prompts to all five.
- Cited by category but not by name (partial): the engine describes what a good provider in your category looks like, or names competitors, but does not mention you even though you are a legitimate option in that market. This usually means a retrieval or entity problem: the engine cannot confidently confirm who you are or is not finding your content at all.
- Competitors cited, you are not (miss): the engine names specific businesses and none of them is you. This is the most actionable outcome, because you can directly compare what your top-cited competitor has (schema, FAQ content, review volume, third-party mentions) that you do not.
- Generic answer with no citations (category too low-authority): the engine gives general advice ("look for licensed, insured providers with good reviews") without naming any specific business at all. This usually means the whole category, in that specific geography, has not built up enough structured, citable content anywhere for the engine to name names, which is actually an opportunity if you move first.
- Mixed pattern (the most common real-world result): your business wins on some prompts and misses on others. Do not average this into a single verdict. A win on the broad category prompt and a miss on the neighborhood-level prompt are both true at once, and the pattern across all five prompts, not any single one, is what actually tells you where the gap is.
Running the same 5 prompts across 3 engines
Different engines pull from different sources and weight different signals, so the same business can show up cleanly in one and disappear in another. Run all five prompts in each of these three engines to get the full picture, not just one.
ChatGPT: use a paid or free account with web browsing/search enabled (look for the globe or "Search" indicator on the response, not a purely training-data answer). ChatGPT leans on Bing for retrieval, so results here are a strong proxy for whether you are indexed and crawlable at all.
Perplexity: run the same prompts in Perplexity, which shows its sources inline as numbered citations you can click through. This is the easiest engine to audit because you can immediately see exactly which pages it pulled from, including whether it used your own site or a directory listing about you.
Gemini: run the prompts again in Gemini. Google's own model draws on a different index and weighting than ChatGPT and Perplexity, and a business that shows up well in one but not the others is telling you something specific about which technical or content gap is engine-specific versus universal.
Do this on a clean session or incognito window where possible, since some of these tools can be influenced by your account's search and browsing history, which is not representative of what a new customer would see.
It is worth also noting where Google AI Overviews and Claude fit into this picture, since business owners ask about both. Google AI Overviews show up embedded directly in regular Google search results for a growing share of local queries, and checking them means simply running your five prompts as normal Google searches and watching for the AI-generated summary box at the top. Claude does not currently offer live web browsing to all users the way ChatGPT and Perplexity do, so it is a less reliable engine for this specific real-time check, though it is still worth testing occasionally as that changes.
Build a tracking spreadsheet
A one-time check tells you where you stand today. A simple recurring spreadsheet tells you whether your fixes are working. Set up one row per prompt per engine per week, and keep it lightweight enough that you will actually maintain it.
- Date — the week you ran the check, so you can see movement over time.
- Engine — ChatGPT, Perplexity, or Gemini.
- Prompt — which of the 5 prompts (or your own variations) you ran.
- Outcome — cited by name, cited by category, competitor cited, or generic/no citation.
- Sources cited — the specific domains or businesses the engine named, so you can track who is winning the citations you want.
- Notes — anything specific worth remembering (a factual error about your business, an outdated address, a competitor you did not expect to see).
What to fix based on which prompts fail
The pattern of which prompts fail, not just the raw count, tells you where the problem actually lives.
If prompt 5 ("what do people say about [business name]") comes back generic or confused even when prompts 1-4 work reasonably well, that is usually an entity and schema problem: the engine can find category-relevant content but cannot confidently tie it to your specific business identity. Check your Organization schema, sameAs links, and NAP consistency.
If prompt 1 and 2 (the broad and comparison prompts) consistently return competitors and never you, while prompt 5 at least recognizes your business exists, that points to a content depth or indexation problem: your site may not be crawlable, deep enough, or structured well enough to surface for broad category queries even though the engine knows you exist.
If prompt 3 (the neighborhood-level prompt) fails while the city-level prompt 1 succeeds, your content is too city-generic. You likely need neighborhood-specific pages or at least neighborhood mentions woven into your existing service pages.
If prompt 4 (the attribute-matching prompt) fails across the board, your content does not clearly communicate the specific differentiators (fast, affordable, licensed, family-owned) that let an engine match you to a specific customer need rather than a generic category.
If every single prompt across every engine returns a generic answer with no business named at all, you likely have a hard technical blocker: check Bing indexation and confirm GPTBot, ChatGPT-User, OAI-SearchBot and PerplexityBot are not being blocked at your CDN or in robots.txt before assuming it is a content problem.
One more pattern worth watching for: if the engine names you but gets a fact wrong (an old address, a service you no longer offer, a phone number that has changed), that is not a visibility problem, it is an accuracy problem, and it is arguably worse than not being cited at all, since it is actively sending customers incorrect information. Fix the source, usually an outdated directory listing or stale schema, as the priority over everything else on this list.
When DIY hits its limit
Five prompts across three engines, run once, gives you a real signal, not a false one. But it is a sample, not a system. It will not tell you how you compare across the 40-plus query variations a real customer base actually uses, it will not check your site against the full list of technical and schema factors that explain why a prompt failed, and it will not give you a structured, prioritized fix list, just an intuition about roughly where the problem might be.
That is the gap between a free DIY check and a full audit. Our Pro Audit Report runs your business through 40-plus real diagnostic prompts across six engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, plus two more), checks every technical and schema factor that explains the results, and delivers a 55-page report with a scored breakdown and prioritized action list within 48 hours, for $19.
Think of the five-prompt check in this article as the smoke test and the Pro Audit as the full inspection. The smoke test is genuinely useful on its own, and if you run it honestly across all three engines, you will walk away knowing more about your AI visibility than the vast majority of local businesses ever bother to find out. But if it turns up a miss or a mixed pattern and you want to know exactly why, down to the specific schema field or content gap, that is where a structured audit earns its keep.
Frequently asked questions
Do I need a paid ChatGPT account to run this check?
No, but you do need web browsing or search enabled on whatever tier you use, since a purely training-data response will not reflect current information about your business. Look for a globe icon, a "Search" label, or visible source citations in the response to confirm it actually searched the web rather than answering from memory.
Why does my business show up in ChatGPT but not Perplexity, or the other way around?
Different engines pull from different underlying indexes and weight signals differently. ChatGPT leans heavily on Bing for retrieval, Perplexity runs its own crawling and ranking layered on multiple sources, and Gemini draws on Google's index and knowledge graph. A gap between engines usually points to something engine-specific, like Bing indexation status, rather than a universal content problem.
How often should I re-run these 5 prompts?
Weekly is enough to see meaningful movement without wasting time, since AI-cited results do not usually change dramatically day to day. Run the check weekly for the first 8-12 weeks after making any fixes, then move to monthly once your results stabilize.
What if a competitor is cited and I have no idea why?
Click through to whatever source the engine cited them from, which Perplexity shows inline and which you can usually find by asking ChatGPT or Gemini a follow-up like "what source did you use for that." Compare that page against your own equivalent page for schema markup, content depth, and whether it directly answers the question you asked.
Is a generic answer with no business named a bad sign?
It depends on context. If your competitors are also absent from that answer, it usually means the category or geography has not built up enough citable, structured content anywhere yet, which is an opportunity to be the first business the engine learns to name. If competitors are named and you are not, that is a direct miss worth fixing.
Can I trust a single run of these prompts, or should I run them multiple times?
Run each prompt at least twice on different days before drawing a firm conclusion. Generative engines are not fully deterministic, so a single run can occasionally miss or include a business by chance. A consistent result across two or three runs on different days is a much more reliable signal.
How is the Pro Audit different from just running these 5 prompts myself?
The Pro Audit runs 40-plus prompt variations (not 5) across six engines (not three), and pairs the prompt results with a full technical and schema audit of your site that explains why each result happened and exactly what to fix, delivered as a scored 55-page report within 48 hours for $19.
Not sure how AI sees your business?
Get your 55-page Pro Audit for $19 — delivered in 48 hours. Shows exactly where you stand in ChatGPT, Perplexity, Google AI Overviews and the Local Pack.
Keep exploring
Related services, industries, cities and resources from Local Visibility AI.