Quick answer: You can't see inside ChatGPT's index, but you can measure the signals that predict AI visibility — Google Search Console's Discover and query data, branded search volume, direct/referral traffic spikes with no matching campaign, and manual spot-checks where you ask AI Overviews, ChatGPT, and Perplexity your own money questions. Track these monthly, not daily; AI citation patterns shift slowly and noisy day-to-day data will mislead you.
Key takeaways
- Google Search Console won't label any click as "from AI Overviews," but a rising ratio of impressions to clicks on question-style queries is a documented signal that Google is answering the query directly instead of sending traffic, according to how Google's own Search Central documentation describes AI Overview reporting.
- Branded search volume — people typing your business name into Google after getting an AI answer — is one of the few proxies you can track for free in Search Console's Performance report.
- No major AI assistant currently offers a public API or dashboard that reports when or how often it cited a specific website, so manual, repeated testing of real queries is still necessary.
- Most small businesses should check AI visibility monthly, not daily — citation patterns in large language models change slowly, tied to retraining and index refreshes, not daily content tweaks.
Step 1: Accept What You Can't Measure Directly
Start here because it saves you weeks of chasing numbers that don't exist. There is no dashboard — free or paid — that shows you every time ChatGPT, Perplexity, or Google's AI Overviews mentioned your business.
These systems don't publish citation logs. OpenAI, Perplexity, and Google don't give website owners a report of "you were cited 47 times this month." That's different from traditional SEO, where Google Search Console has always told you exactly which queries showed your site and how often.
This matters because plenty of tools marketed as an ai seo tool will show you a number that looks precise — "AI visibility score: 82" — built on sampled test queries, not a real feed of assistant behavior. Treat those scores as directional estimates, not facts. The honest starting point is: you're measuring proxies, not ground truth, and every method below is a proxy.
Step 2: Pull the Right Signals From Search Console
Google Search Console is still your best free source, but you have to look at different metrics than you did five years ago. The old playbook — clicks and average position — misses what's happening when Google answers a question directly on the results page instead of sending someone to your site.
Here's what to actually watch:
- Impressions with falling or flat clicks on question-phrased queries ("how much does X cost," "is X worth it"). This pattern often means Google is generating an AI Overview or featured snippet from your page's content without the user needing to click through — your content is being used, just not visited.
- Click-through rate (CTR) by query, not just by page. A page that ranks #3 but shows near-zero CTR on informational queries is a candidate for AI summarization eating the click.
- Query length and phrasing shifts. If you're newly showing impressions for long, full-sentence queries you never targeted, that's often AI-generated query fan-out — Google or an assistant rephrasing a user's question into several searches behind the scenes.
- Discover and "Web" filter comparisons. Check whether impressions moved in the plain web results versus Discover; a jump in one without the other tells you which surface is picking up your content.
We covered how to turn these same Search Console patterns into actual topic ideas in Search Console Blog Topic Ideas: A Step-by-Step Guide — worth pairing with this tracking work since the two feed each other.
Don't skip this: a query showing high impressions and near-zero clicks isn't a failure — it can mean your content was good enough to get summarized. Don't panic and rewrite the page; check first whether it's still driving branded searches or direct visits downstream.
Step 3: Run Manual Spot-Checks on Real Queries
The only way to actually see whether your business shows up in a generative answer is to ask the assistant yourself, regularly, using the exact questions your customers would type. This is tedious but it's the ground truth no tool fully replaces.
Build a list of 15–20 real questions your customers ask before buying — not your brand name, the underlying question. A plumber might use "how much does it cost to fix a running toilet" or "do I need a permit to replace a water heater." Run each one through:
| Where to check | What to look for | How often |
|---|---|---|
| Google AI Overviews | Your domain named or linked in the summary box | Monthly |
| ChatGPT (with browsing/search on) | Your business named, described, or linked in the response | Monthly |
| Perplexity | Citation footnotes pointing to your site | Monthly |
| Google featured snippets | Your page pulled into the boxed answer | Monthly |
Keep this in a simple spreadsheet: query, date, which assistant, cited or not, exact phrasing used. Do it the same week each month so you're comparing apples to apples. We go deeper on interpreting these results in How to Track Whether AI Tools Are Recommending Your Business.
Step 4: Track Branded Search and Direct-Traffic Lift
Branded search — people googling your business name — is one of the most underused signals for AI visibility, and it's sitting in your Search Console data for free right now. When an AI assistant recommends a business by name without a clickable link, the natural next move for a curious user is to type that name into Google.
To check this:
- Filter the Search Console Performance report to queries containing your exact business name.
- Compare month-over-month volume, not day-to-day — a single viral mention creates noise, a real trend takes weeks to show.
- Cross-reference with your analytics' direct traffic and "unknown referrer" traffic, which often spikes when someone copies a name mentioned in a chat response rather than clicking a link.
- Rule out obvious causes first — a new ad campaign, a press mention, a seasonal spike — before crediting AI assistants.
This won't tell you which assistant sent the person. But a sustained rise in branded search with no matching campaign, ad spend, or press hit is a real, attributable signal that something outside your direct control is putting your name in front of people.
Step 5: Use Citation-Monitoring Tools to Save Time, Not Replace Judgment
A handful of tools now automate the manual spot-check from Step 3, running your target queries against multiple AI assistants on a schedule and flagging when your domain appears. That's useful — it turns a monthly chore into a dashboard.
What to know before you pay for one:
- They're sampling, not surveillance. No tool sees every real user conversation with ChatGPT; they run their own test queries on a schedule, same as you could manually.
- Coverage varies by assistant. Some tools track Google AI Overviews well but have thin coverage of Perplexity or Microsoft Copilot — check exactly which assistants and how often before subscribing.
- Free options exist. Manually running your own query list monthly costs nothing but time, and gives you full control over which questions matter to your actual customers.
- Paid tools earn their keep at scale. If you're tracking 50+ queries across multiple assistants weekly, automation pays for itself in time saved; for a 15-query monthly check, it may not.
Whichever route you pick, the metric you're ultimately watching — being named in an AI answer — depends on the same underlying thing: content structured so an assistant can lift a clear, self-contained answer from it. We break down exactly how to write for that in How to Write Blog Content That Ranks and Gets Cited by AI.
Step 6: Focus on the Three Metrics That Actually Move Together
Of everything above, three metrics correlate most consistently with real AI visibility, and they're worth checking as a monthly ritual rather than a one-time audit.
- Impression-to-click ratio on question queries in Search Console. A widening gap here — impressions climbing while clicks stay flat — is the clearest free signal that your content is being pulled into a summarized answer.
- Branded search volume trend. Rising name searches with no paid or PR driver behind them suggest people are hearing about you somewhere you don't directly control.
- Manual citation rate on your own test query list. Out of your 15–20 real customer questions, what percentage returned your business by name last month versus this month? That trend line, tracked consistently, is more honest than any single "visibility score."
Watch these three together, monthly, for at least a quarter before drawing conclusions. Generative engines don't update their sense of "who's trustworthy on this topic" overnight — it tracks more closely with sustained, consistent publishing than with any single optimized page. That's also the logic behind why some businesses get recommended by AI and others don't: consistency and topical depth over time, not a one-off trick.
What to Do Next
Set a recurring monthly calendar block — 30 minutes is enough — to pull your Search Console question-query report, run your test-query spreadsheet, and check branded search volume. Do it the same week every month so the trend lines mean something.
If you're not publishing often enough to have fresh content for assistants to find in the first place, the measurement step is premature — you need something to measure. Segeo publishes a new, on-topic blog post to your site every day, pulled from live search and industry trend data rather than a recycled calendar, and includes AI-visibility tracking built into the monthly plan so you're not stitching this together by hand. Plans start at $89/month.
