How to Track AI Search Rankings (ChatGPT, Copilot & More)

You cannot "rank track" AI search the way you rank track Google. There is no stable position 1 to 10 inside a ChatGPT or Copilot answer. Ask the same question twice and you can get two different shortlists. So the whole category of "Copilot rank trackers" is selling you a number that does not really exist.

Here is what actually works. Stop chasing a phantom ranking and measure the funnel instead: how often AI surfaces you (impressions), how much traffic it sends you (referral), and how often it names you over competitors (citation share). Two of those three layers are now free and native inside Google Search Console and GA4, both shipped in the last month. You only need to pay for the third.

I track AI referral across 100 ecommerce brands at StudioHawk. This is the exact framework I use, the free tools that cover most of it, and the two paid tools worth the money.

Why "rank tracking" is the wrong frame for AI search

Traditional rank tracking works because Google returns a roughly stable, ordered list. Position 3 today is usually position 3 tomorrow. AI answers break that assumption completely.

AI responses are generated, not retrieved. The model synthesises an answer on the fly, pulls different sources depending on phrasing, personalises by account and history, and rewords itself every time. There is no fixed slot to "rank" for. A brand that gets named in an answer this morning can vanish from the same prompt this afternoon.

So when a tool promises your "Copilot ranking," ask what it is really measuring. Almost always it is citation frequency: across a set of prompts run repeatedly, how often does your brand get mentioned. That is a useful metric. It is not a ranking, and treating it like one leads to bad decisions.

The frame that holds up is full-funnel measurement, from impressions through to revenue. Here is what each layer tells you and where to get it.

Track AI impressions free with Search Console's new gen-AI report

This is the big one, and it is days old. On 3 June 2026 Google shipped dedicated Generative AI performance reports inside Search Console. For the first time you can see, natively, how often your URLs appear inside AI Overviews and AI Mode, plus generative AI features in Discover.

What it shows: impressions and the pages that surfaced, broken down by country, device and date. What it does not show yet: clicks. It is impressions-only at launch, the data starts from 18 May 2026 with no history, and it is rolling out to a subset of sites first. There is also an opt-out toggle that pulls your content from AI Overviews and AI Mode without touching your organic rankings, and Google starts acting on that setting from 17 June 2026.

If you have the report in your account, this replaces guesswork. You no longer need a third-party tool to know whether Google's AI features are surfacing your pages. Check it first, before you pay anyone.

Track AI referral traffic in GA4 (and the gotcha that loses you half of it)

The second free layer landed on 13 May 2026, when Google added a native AI Assistant channel to GA4's default channel group. Sessions arriving from a recognised AI domain now get the medium ai-assistant and land in their own channel automatically. No setup required.

Here is the catch that no vendor listicle mentions, because it makes the free option look good. The native channel only recognises a short list: ChatGPT, Gemini and Claude. Perplexity and Copilot are not captured. They fall back into Referral. And on top of that, Statcounter data from March 2026 found that between 35% and 70% of AI referral sessions arrive with no referrer at all, so they get dumped into Direct.

Translation: if you rely on GA4's native channel alone, you are seeing maybe half of your real AI traffic, and you are blind to Copilot entirely.

The fix is a custom channel group with a regex that catches every engine. This is the source-match pattern I use:

chatgpt\.com|chat\.openai\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|bing\.com/chat|deepseek\.com|grok\.com|meta\.ai|you\.com

Build that as a custom channel group in GA4, and the Direct-traffic leakage is a separate fight: append utm_source where you can, and watch the patterns I break down in my guide to utm_source=chatgpt.com and tracking ChatGPT traffic in GA4.

How Copilot actually works (why a good Bing rank is not enough)

Most "Copilot rank tracker" articles tell you Copilot "grounds its answers in Bing" and stop there. That sentence is doing a lot of unexplained work, and the gap is where people waste money.

Copilot is embedded directly in Windows, Edge and Microsoft 365. It reaches buyers at their desk before they ever open a browser tab. When it answers, it does pull from Bing's index, but it then synthesises a conversational response and chooses which sources to cite. Ranking well in Bing gets you into the candidate pool. It does not guarantee Copilot names you in the answer.

So tracking your Bing position is a weak proxy for Copilot visibility. You can sit at Bing position 2 and never get cited, while a page at position 6 with cleaner, more extractable content gets named every time. If Copilot matters to your buyers, optimise for citation, not just for the blue-link rank. I cover the mechanics in Copilot search optimisation and the broader engine differences in Google vs Bing.

One example of why this is worth the effort: an enterprise conference once booked me as a speaker, over 400 attendees, through my website inquiry form. I had no idea where they found me. When I asked, they had asked Microsoft Copilot about AI search experts and my name came up. That is a booking I can trace directly to a Copilot citation, not a Bing ranking.

What AI referral actually looks like (benchmarks from 100 brands)

The honest problem with this whole topic is that almost nobody publishes real numbers. The listicles quote Microsoft's earnings deck. Here is first-party data instead, from the 100 ecommerce brands I track at StudioHawk.

• ChatGPT referral grew 19x year over year across the portfolio.
• That traffic drove roughly $690,000 in tracked revenue from 340,000 sessions.
• 91 of the 100 brands now receive AI referral traffic that did not exist 18 months ago.
• Aggregate conversion rate sat around 0.9%, with huge variance by category.

For context on share, ChatGPT is still the dominant AI referrer at roughly 78% of AI traffic, and Gemini overtook Perplexity as the number two source in early 2026. So if you only set up tracking for one engine, make it ChatGPT, then add the rest with the regex above.

Use those figures as a rough benchmark. If AI referral is a meaningful and growing slice of your sessions, you are in line with the market. If it is still zero, that is the gap to close, and the measurement comes first.

When you actually need a paid tool (Peec and Profound)

Everything above is free. The one thing GSC and GA4 cannot show you is layer 3: how often AI names you versus your competitors across a set of buying-intent prompts. That is the genuine job of a paid AI visibility tool, and it is the only reason to pay.

Two I rate: Peec AI and Profound. Both run repeated prompt sets across multiple engines and report citation share, competitor comparison and which sources the models lean on. That competitor view is the bit that earns its keep, because seeing that a rival gets named in 8 of 10 answers while you get named in 2 is a far more useful signal than your own number in isolation.

What the category mostly gets wrong:

• Selling "rank" as if it is stable. It is citation frequency. Treating a noisy daily number as a ranking leads to panic over normal variance.
• Vendor-biased listicles. Most "best Copilot tracker" posts are written by a tool that conveniently ranks itself number one. Read them knowing that.
• Charging for the free layers. Plenty of tools repackage impression and referral data you can now get from GSC and GA4 for nothing.
• Engine-count theatre. "We track 9 engines" means little if the citation data is not actionable. Five engines of useful data beats nine of noise.

My rule: most small and mid-size businesses need GSC plus GA4 and nothing else. Add Peec or Profound when citation share-of-voice against named competitors becomes a metric your leadership team actually asks about.

The setup, in order

1. Check the GSC Generative AI report for AI impressions. Free, native, start here.
2. Confirm the GA4 AI Assistant channel is showing, then build the custom channel group with the regex so you capture Copilot, Perplexity and the rest.
3. Tighten Direct leakage with utm tagging where you control the link.
4. Only then, if competitor citation share matters, add Peec or Profound.

That order saves most teams a subscription they do not need, and gives you a measurement stack that reflects how buyers actually find you across AI search. If you want a hand setting it up across a brand portfolio, that is the work I do every day. See how I track it on the AI search tracker and how to measure AI search traffic.

Frequently asked questions

Can you track Copilot rankings for free?

You cannot track a fixed Copilot "ranking" because AI answers have no stable positions. You can track the things that matter for free: AI impressions in Search Console's Generative AI report, and Copilot referral traffic in GA4 once you add a custom channel group, since Copilot is not in GA4's native AI channel.

Does GA4 track ChatGPT and Copilot traffic automatically?

Partly. Since 13 May 2026 GA4 has a native AI Assistant channel, but it only recognises ChatGPT, Gemini and Claude. Copilot and Perplexity fall into Referral, and 35% to 70% of AI sessions land in Direct with no referrer. You need a custom channel group with a regex to capture all engines.

What is the best ChatGPT source tracking tool?

For traffic, GA4 with a custom channel group is the best source and it is free. For citation share against competitors, Peec and Profound are the two paid tools worth using. Start free, add a paid tool only when competitor citation data becomes a business metric.

Why does my AI traffic show up as Direct?

Many AI clients strip the referrer when they send a click, so GA4 has nothing to attribute and files the session under Direct. Statcounter put this at 35% to 70% of AI sessions in March 2026. A regex-based channel group catches what referrers it can, and utm tagging on links you control recovers more.

Is tracking my Bing ranking enough for Copilot?

No. Copilot grounds in Bing's index but then synthesises and chooses which sources to cite, so a strong Bing rank only gets you into the candidate pool. You can rank well in Bing and never be named by Copilot. Track citation, not just the Bing position.