AI Search Tracking and Analytics: 3 Free Ways

Adam Khaled
Founder, Web25

Adam Khaled
Founder, Web25
Quick answer: AI search tracking, also called AI search analytics, means checking how often ChatGPT, Google’s AI features, Perplexity, and Microsoft Copilot mention or link to your business. To track AI search visibility for free, use Search Console, Bing Webmaster Tools, and GA4, plus a monthly prompt check run several times per prompt. A paid tool pays off once you need more prompts than you can run by hand.
ChatGPT rarely lists the same brands in the same order twice, so tracking your AI visibility works differently from the rank tracking you may already know. What you can measure is how often you appear, and where the visits come from.
AI search tracking is the routine of measuring whether AI tools mention your business, cite your pages, or send visitors to your site. When you turn those numbers into a monthly report, people call it AI search analytics. It gives you three signals to watch. They are impressions inside Google’s AI features, visits that arrive from AI assistants, and how often a chatbot names you when someone asks a question you’d want to answer.
A mention means the AI names your business in its answer. A citation means it links to one of your pages as a source. You can get one without the other, so log them separately. The broader idea behind all of this is covered in our guide to AI visibility. The wider practice is called answer engine optimization, and tracking is how you check it’s working.
To track AI search visibility, check what Google and Bing report about their own AI features, check your AI assistant visits in GA4, then run a fixed set of prompts several times and log the results.
Search Console’s Generative AI performance report shows how many times your pages appeared in Google’s AI Overviews and AI Mode. You can split it by page, country, device, and date. It reached all websites worldwide on August 31, 2026, according to Google’s help page.
The report has limits worth knowing before you build anything on it. It shows impressions only, with no clicks. It leaves out Search Labs experiments, and a site needs enough AI impressions before the report appears at all. Google defines impressions as “how many times links to your site were shown to a user in a generative AI feature on Google Search.” The full walkthrough is in our guide on how to measure AI Overview traffic.
Bing Webmaster Tools has a matching report. Its AI Performance report shows how your content appears in AI answers “across Microsoft Copilot, Bing, and select partner AI experiences,” according to Microsoft. It launched as a public preview in February 2026, and on June 16, 2026 Microsoft added Intents, Topics, Citation Share, and Compare in preview. Citation Share is your share of the citations for one grounding query. Microsoft doesn’t name the partners, so use the prompt check in Method 3 for ChatGPT and Perplexity.
Open Reports, then Acquisition, then Traffic acquisition, and look for the AI Assistant row in the default channel group. Google added that channel on May 13, 2026, and it assigns the medium “ai-assistant” when the referrer matches a recognized assistant. Google’s Analytics release notes name ChatGPT, Gemini, and Claude as examples.
The channel only catches visits that carry referrer information. Visits from in-app browsers, mobile apps, and pasted links can still land in Direct, so your real AI traffic can be higher than the channel shows. Seer Interactive’s tracking guide notes that ChatGPT adds utm_source=chatgpt.com to some links, not all. If the channel is missing on your property, search the Session source column for “chatgpt” and the other assistant names.
GA4 also added a Source Group field on June 11, 2026. Google’s release notes say it includes “built-in grouping for emerging traffic sources like ChatGPT (OpenAI) and Perplexity” and that it is populated retroactively, so it can show ChatGPT and Perplexity visits from before May 13, 2026. The AI Assistant channel itself only works forward from its launch, according to Seer Interactive, so older AI visits stay under Referral or Direct.
Run the same set of customer questions through each AI tool several times, and log whether your business appears. A single run is not enough. SparkToro’s study, published January 2026, ran 12 prompts through ChatGPT, Claude, and Google’s AI a combined 2,961 times. The chance of getting the same list of brands twice was under one in 100, and the same order was closer to one in 1,000.
The same study found that visibility percentage across many repeated runs is a reasonable metric. It also warned that any tool promising a ranking position in AI is unreliable. So the method below counts appearances, not positions.
Use the wording customers actually use, such as “who builds websites for contractors in Ohio” or “how much does website maintenance cost”. Pull them from sales calls, emails, and the questions people ask before they book. Prompts you invent from your own service list will miss how buyers really ask. Mix three kinds of question: discovery (who does this), comparison (which option is better), and decision (is this one worth it). Once the list is set, keep it as it is.
Open a fresh chat for every run so earlier answers don’t shape the next one. Five runs per prompt per platform is a workable starting point. One run tells you almost nothing, for the reason SparkToro’s data shows above. Use the same settings every month (logged out or the same account, same location, same mode) and write them down. Answers change between runs anyway, so you want one less thing changing.
For every run, record whether your business was named, whether a page from your site was linked, and whether what the AI said about you was correct. Keep the raw notes. The wrong-answer notes are often the most useful, because they point at pages that need fixing. If the AI got you wrong, here is how to fix what ChatGPT says about your brand. You can copy this log layout into a sheet.
| Date | Platform | Prompt | Run # | Business named? | Page linked? | Correct? | Competitors named |
|---|---|---|---|---|---|---|---|
| Example: month 1 | ChatGPT | Who fixes storm-damaged roofs in Columbus? | 1 | Yes (Example Roofing) | No | Yes | Roofer A, Roofer B |
The row above is an example with placeholder values, not real data. Add one row per run.
Divide the runs where you appeared by the total runs. If you appeared in 12 of 60 runs, your visibility is 20 percent. For a simple share of voice, count every brand mention in your log, yours and your competitors’, and divide your mentions by that total. If you were named 9 times out of 30 brand mentions, your share of voice is 30 percent. Both numbers are examples. Rerun the same prompts next month and compare. Change the prompt list rarely, or the comparison stops meaning anything.
Perplexity and ChatGPT often answer the same question differently, so keep separate columns for each. Our comparison of ChatGPT and Perplexity citations explains why.
Start with Google Analytics 4 if you only have time for one, because it needs no setup and shows real visits. Add the manual prompt check next, since it’s the only one of these free methods that checks ChatGPT and Perplexity answers themselves. The table shows what each method covers.
| Method | What it shows | What it misses | Time each month |
|---|---|---|---|
| Search Console Generative AI report | Impressions in Google AI Overviews and AI Mode, by page, country, and device | Clicks, and every non-Google AI tool | About 10 minutes |
| Bing Webmaster Tools AI Performance | Citations of your pages in Microsoft Copilot, Bing, and select partner AI experiences | AI tools outside Copilot, Bing, and Microsoft’s unnamed partners | About 10 minutes |
| GA4 AI Assistant channel | Sessions from recognized AI assistants that pass a referrer | Visits with no referrer, and answers where nobody clicked | About 10 minutes |
| Manual prompt check | Whether ChatGPT, Perplexity, and others name or link your business | Prompts you didn’t think of, and it takes the most effort | 1 to 2 hours for 10 to 15 prompts |
The time estimates are our own rough figures for a small site, not measured benchmarks.
An AI search analytics report needs five numbers, and each one has a free source. Google, Bing, and GA4 cover the first three. Your prompt log covers the last two.
Report the same five numbers every month and compare them with the previous month. Never judge from a single run, because the same prompt rarely returns the same list of brands twice.
A paid tool is worth it when you need to track more prompts than you can run by hand, or compare yourself against competitors every week. Semrush’s AI Visibility Toolkit reports an AI Visibility Score, share of voice, and sentiment across ChatGPT and Google’s AI Mode. Ahrefs Brand Radar tracks mentions, citations, and share of voice across AI Overviews, Gemini, Perplexity, ChatGPT, Copilot, AI Mode, and Claude. Some teams now hand this job to an AI agent, which is one use of agentic SEO.
These four paid trackers publish different details about coverage, update speed, and prompt limits. Here is what each vendor’s own page says.
| Tool | AI platforms the vendor names | Prompts in the entry plan | How often it updates | Says how many runs per prompt? |
|---|---|---|---|---|
| Semrush AI Visibility Toolkit | Platforms like ChatGPT and Google AI Mode | 25 tracked prompts | Prompt Tracking daily, Brand Performance weekly | Not stated on the page |
| Ahrefs Brand Radar | AI Overviews, AI Mode, Gemini, Copilot, Perplexity, ChatGPT, plus Claude on custom prompts | 5 custom prompts (Lite). The prompt index built from Ahrefs keyword data is priced separately | Custom prompts daily, weekly or monthly | Not stated on the page. Each prompt uses 1 check per platform, location and update (Claude uses 8) |
| Otterly.AI | ChatGPT, Google AI Overviews, Perplexity, Copilot. Claude, Google AI Mode and Gemini are extra add-ons | 15 prompts (Lite plan) | Daily | Not stated on the page |
| Advanced Web Ranking | Google AI Mode (170+ countries), ChatGPT (54), Perplexity (13), Gemini (4) | Uses keyword units, and AI Mode uses twice the units of standard engines | After each scheduled update or on demand | Not stated on the page |
This is what the vendor pages listed when we checked in October 2026, and vendors change them often, so we left prices out. Our guide to GEO tools and agencies has a longer list.
Before you pay, ask how many times the tool runs each prompt. None of the four vendor pages above says how many times it runs a prompt in each update. Ahrefs says each prompt uses 1 check per platform, location and update, and Claude uses 8 checks per update. One run per prompt would give you noise. For a business with one location and a few services, the three free methods above are usually enough.
Web25’s AEO and GEO plans include a monthly plain-English report, and the Growth plan adds competitor visibility tracking. We are the agency that publishes this guide, so weigh that. You can run all three free methods yourself, and this article gives you everything you need to start. If you’d like help, see our AEO and GEO service page.
Rank tracking records one position for one keyword. AI tools give no fixed position, and the same prompt returns a different list of brands almost every time. So AI search tracking records how often your business appears across repeated runs of the same prompts, and whether the answer links to your site. It measures frequency, not a rank.
Yes, when the visit carries referrer information. GA4’s AI Assistant channel groups recognized chatbots such as ChatGPT, Gemini, and Claude automatically. Visits from in-app browsers, mobile apps, or pasted links can still land in Direct. If the channel is missing on your property, search the Session source column for the assistant’s name instead.
No. The Generative AI performance report only covers Google’s own AI Overviews and AI Mode. ChatGPT, Perplexity, and Claude never appear in it. For those platforms you need referral data in Google Analytics, a manual prompt check, or a paid tool that runs prompts against each platform for you.
Start with 10 to 15 that mirror real customer questions. That’s a practical starting size for a small business, not a researched rule. It’s small enough to run by hand each month and large enough that one odd answer won’t skew the percentage. Add prompts only when a new service or location needs its own coverage.
Once a month is enough for most small sites. AI answers change from run to run, so daily checks mostly record noise. Compare each month’s visibility percentage against the previous one and look for a steady direction over a quarter. Check sooner only after a real change, such as a new service page or a site migration.
Usually the free methods are enough. A business with one location and a handful of services can cover its main customer questions with 10 to 15 prompts and a monthly GA4 check. A paid tool starts to make sense across several locations or competitors, where hand-running prompts takes hours every month.
An AI search tracker is a tool or routine that reruns a fixed set of prompts on AI platforms and records whether your brand is named or linked. The free version is a manual sheet you update each month. Paid tools automate the same job. Before you buy one, ask how many times it runs each prompt, because a single run is mostly noise.
Partly. GA4’s AI Assistant channel only labels visits from its May 13, 2026 launch forward, so older AI visits stay under Referral or Direct. The newer Source Group field is populated retroactively and groups sources like ChatGPT and Perplexity, so it works on historical data. A Session source search for chatgpt also finds some older visits.
Open Google Analytics 4 today and check the AI Assistant row under Traffic acquisition, then write your first 10 prompts. If you run a local business, read our guide to AEO for local business next, or book a discovery call to talk through your own site.
Written by Adam Khaled, Founder, Web25. Last updated October 2026.

Odoo Shopify integration in 7 steps: check your Odoo plan and hosting, pick an Odoo module, a Shopify app or custom code, and sync stock across locations.

QuickBooks Shopify integration in 7 steps: pick the right QuickBooks, choose an app or custom code, and make each Shopify payout match your bank deposit.

ERP website integration explained: what your website can pull from or send to your ERP, five ways to connect them, and 6 steps to scope the job.