Quick answer: Track AI visibility by running the same set of real customer questions through ChatGPT, Perplexity, and Google AI Overviews every week, logging whether your business shows up, and noting what sources those tools cite when it doesn't. There's no official dashboard for this yet — no "Search Console" for ChatGPT — so it's a manual process of asking, recording, and comparing over time. Most businesses can set up a workable tracking system in under an hour.
Key takeaways
- No AI platform currently offers a public analytics dashboard showing when your business was recommended, so tracking means manually testing a fixed list of prompts on a regular schedule.
- Perplexity and Google AI Overviews show their source links directly in the answer, which makes them easier to audit than ChatGPT, which often doesn't cite sources at all.
- The businesses AI tools recommend tend to have consistent, structured information across their website, Google Business Profile, and third-party sites — not necessarily the biggest ad budget.
- Checking once isn't tracking — AI answers change as models update and as new content gets published, so a single test only tells you where you stand today, not whether you're improving.
Step 1: Build Your Prompt List Before You Test Anything
Start by writing down 10 to 15 questions your actual customers would type into ChatGPT or ask Google, not questions about your business by name. The goal is to see if you show up when someone doesn't already know you exist.
Good prompts sound like real searches:
- "Best [your service] in [your city]"
- "Who should I hire for [specific problem you solve]?"
- "What's a good [your service] near [your city] for [specific situation]?"
- "How much does [your service] typically cost?"
- "[Your industry] recommendations for [specific customer need]"
Skip prompts like "tell me about [your business name]" — that only tests whether the AI has heard of you, not whether it recommends you over competitors. The whole point of tracking is to see how you rank in the eyes of the algorithm when you're not the one bringing your own name up.
Keep this list in a spreadsheet. You'll reuse the exact same prompts every time you test, so the comparison is apples to apples.
Step 2: Run the Same Prompts Across Multiple AI Tools
Test each prompt in at least three places: ChatGPT, Perplexity, and Google's AI Overviews (the AI-generated summary that appears above regular search results). Each one pulls information differently, so showing up in one doesn't mean you'll show up in another.
| Tool | Shows sources? | How to check it |
|---|---|---|
| Google AI Overviews | Yes, linked citations | Search your prompt on Google, read the AI box at the top |
| Perplexity | Yes, numbered citations | Ask the prompt directly in Perplexity's chat |
| ChatGPT | Sometimes, depends on mode | Ask in a fresh chat; use "search the web" mode if available |
Perplexity is the easiest to audit because it shows exactly which pages it pulled from for every claim — we covered how it decides what to cite in Perplexity AI for Business: What It Is and Why It Matters. ChatGPT is the trickiest since it doesn't always browse the web or show sources, so its answer reflects whatever it learned during training plus whatever it retrieves in that session.
Log three things for each tool: whether you were mentioned, where you ranked if there was a list, and which competitors showed up instead of you.
Step 3: Read the Citations, Not Just the Answer
When an AI tool does cite sources — Perplexity and AI Overviews both do this — that citation list is more useful than the answer itself. It tells you exactly which pages the model trusted enough to pull from, which means you now know exactly what to improve.
Say you search "best plumber for tankless water heaters" and a competitor's name comes up. Click into the cited source. It's often one of a few things:
- A blog post that directly answers the question with specifics (brand names, price ranges, install times)
- A review site or directory listing with detailed, structured information
- A Reddit thread or forum where someone recommended them by name
- Their own website, if it has a clear page answering that exact question
If your competitor's blog post is getting cited and you don't have an equivalent page on your own site, that's your gap. This is the same logic behind how ChatGPT decides which businesses to recommend — it favors content that answers a specific question clearly, not vague "about us" pages.
Step 4: Set a Repeatable Schedule, Not a One-Time Check
Test your prompt list on the same schedule every time — weekly or biweekly works for most small businesses — because AI answers shift as models update and as new content gets indexed. A single check tells you where you stand today. Only repeated checks tell you if you're actually gaining ground.
Don't skip this: Testing once and calling it done is the most common mistake in AI visibility tracking. These models get updated, retrained, and re-indexed constantly, so a business that showed up in March can disappear by June with no warning and no notification — you'll only know if you're still checking.
Keep your log simple. A spreadsheet with columns for date, prompt, tool, mentioned (yes/no), rank position if applicable, and cited source is enough. Don't overbuild this — the value is in the consistency of checking, not the sophistication of the tracker.
Step 5: Compare Against Your Traditional Search Visibility
Check whether your AI visibility tracks with your regular Google rankings, because the two don't always move together. A business can rank well in traditional search results and still get skipped by AI Overviews, or vice versa — the ranking signals overlap but aren't identical.
If you already track keyword rankings, run the comparison side by side:
| Signal | Where to check it | What it tells you |
|---|---|---|
| Traditional Google ranking | Google Search Console, manual search | Whether your page ranks for the keyword itself |
| AI Overview appearance | Manual search, look for the AI box | Whether Google's AI trusts your page enough to summarize it |
| ChatGPT/Perplexity mention | Manual prompt testing | Whether AI tools recommend you without you being searched by name |
A mismatch is informative. Ranking well on Google but never showing up in AI Overviews for the same query usually means your page ranks on authority signals (backlinks, domain age) but doesn't actually answer the question in a way the AI can summarize cleanly. That's a content structure problem, not a visibility problem — and it's fixable by rewriting the page to directly answer the question in the first sentence or two.
If you're not sure whether the gap is about content or about being findable at all, our post on 5 signs your business is invisible to AI assistants walks through how to tell the difference.
Step 6: Fix What the Citations Are Telling You
Once you've logged a few weeks of data, act on the pattern instead of chasing every individual answer. AI tools reward clear, specific, structured content — this is the practice generally called generative engine optimization, or writing content specifically so AI systems can find, understand, and cite it.
Common fixes based on what the tracking usually reveals:
- You're not mentioned anywhere. Your website likely doesn't have a page that directly answers the customer question in plain language — start there before anything else.
- You're mentioned but ranked below competitors. Check what their cited page has that yours doesn't — specifics like pricing, service area, or a direct answer in the first sentence.
- You show up in Perplexity but not ChatGPT. This is normal since ChatGPT relies more on training data and less on live web browsing by default — consistency over time is what eventually shows up in training data.
- Your Google Business Profile info doesn't match your website. Inconsistent hours, addresses, or service descriptions across platforms make AI tools less confident citing you at all.
What to Do Next
Set up your prompt list this week, run it once to get a baseline, then put a recurring reminder on your calendar to re-test in two weeks. That single habit — ask, log, compare — is 90% of what "AI visibility tracking" actually means right now, since no platform has built a proper analytics tool for it yet.
If the gap you find is a content gap — no page on your site actually answers the questions your customers are asking AI tools — that's the fastest lever to pull, and it's also the one most businesses let slide because writing a genuinely useful page every week is a lot of ongoing work. That's the specific problem Segeo is built to solve: it publishes a new, on-topic blog post to your site every day, picks topics from real search and trend data instead of a recycled content calendar, and tracks your AI visibility as part of the plan — so you're not doing steps 1 through 6 by hand every single week.
