How to Build an AI Overviews Tracker Using Semrush and Search Console
Key takeaways
|
Search "AI Overviews tracker" and you'll get a page of listicles comparing SaaS tools. Neil Patel has one. Most of the AI visibility vendors have one. They're fine if you're buying.
This post is for the team that already pays for Semrush, already has Search Console, and would rather spend an afternoon building the tracker than another line item buying it. After more than 16 years of stitching marketing data together, I've found the homegrown version answers the questions that actually matter, and the SaaS version answers the questions the vendor's dashboard designer thought of.
Two things changed in 2026 that make this possible. Google Search Console now reports AI Overview and AI Mode impressions as a separate view, rolled out globally on August 31. And Semrush has tracked AI Overviews as a SERP feature for long enough that the data is stable.
What does an AI Overviews tracker need to measure?
Four things. Which of your target keywords trigger an AI Overview. Whether your domain is cited inside that Overview and in which slot. Which of your pages are earning impressions in Google's generative AI features. And how each of those changes over time. Rankings are a fifth, optional column for context, not a proxy. |
Most SaaS trackers answer the first two well and the third not at all, because the third comes from Search Console and requires your property access. The homegrown version has an edge here.
The order matters. Trigger rate tells you where the surface exists. Citation tells you whether you're on it. Impressions tell you how often that's happening in the real world rather than in a rank tracker's sample. Trend tells you whether the work is working. If you only track rankings, you'll miss all four, which is the whole argument in citation rate as the KPI.
What does Google Search Console show for AI Overviews?
Since June 2026, and globally since August 31, 2026, Search Console has a generative AI performance report under Performance that shows impressions from AI Overviews and AI Mode combined, broken down by page, country, device, and date. It does not show clicks, CTR, queries, or position. It's a visibility signal, not a traffic signal. |
Open Search Console, go to Performance, and look for the generative AI view. If it's not there, one of three things is true: your property isn't in the rollout yet, your site hasn't received enough AI impressions to populate the report, or someone turned on the Search generative AI opt-out under Settings. Check the third one first. I've seen a security team flip it without telling marketing, and it's the same failure pattern as Cloudflare's AI crawler default.
What you get is a page-level impression count for AI surfaces. Export it weekly. The pages that earn AI impressions are your citation winners on Google, and comparing that list to your organic click leaders is the fastest way to see the ranking vs citation gap on your own site.
What you don't get is the query. Google hasn't exposed prompt-level data, and there's no timeline for clicks. So Search Console tells you which pages, not which questions. For the questions, you need Semrush.
How does Semrush track AI Overviews?
Semrush records AI Overviews as SERP feature code 52. In Position Tracking, any tracked keyword that triggers an Overview is flagged, and the report shows which domains are cited inside it. In Keyword Overview and the API, the same feature flag appears in the SERP features column. The position shown for an AI Overview row is the citation slot inside the Overview, not organic rank. |
That last sentence trips people up. When Semrush shows your domain at position 2 on an AI Overview row, it means you're the second cited source inside the box, not that you rank second organically. Two different numbers. Track both if you want context, but don't confuse them.
Three ways to pull it:
Position Tracking. Set up a campaign with your 20 to 40 target keywords, enable SERP features, and filter to AI Overview. You'll see trigger status and citation status per keyword, per day. This is the cleanest ongoing feed.
Keyword Overview. Check individual keywords for the AI Overview flag and see the cited domains. Good for research, tedious for tracking.
API. If you have API access, the phrase_organic report returns SERP composition including feature 52 rows, and domain_organic with a feature filter returns your keywords that trigger an Overview. I use this for the cannibalization checks and the SEO gap analysis workflow, and it's the most flexible input for the sheet.
Semrush's own AI Toolkit extends this to ChatGPT and Perplexity with prompt-based tracking, which is a separate product and a separate budget. This post stays on the Google side.
How do you build the tracker?
One sheet, four tabs. Tab one is your keyword list with volume, KD, and intent. Tab two is the weekly Semrush pull: keyword, AI Overview triggered yes or no, your domain cited yes or no, citation slot, organic rank. Tab three is the weekly Search Console export of AI impressions by page. Tab four joins them on page URL and rolls up trigger rate, citation rate, AI impressions, and week-over-week change. |
Tab 1, Keywords. Your target set. Twenty to forty is plenty. Mix informational, comparison, and category terms. Add the URL you intend to win each with. Pull volume and KD from phrase_these and phrase_kdi (two separate calls, they can't be combined). This is the same list you'd build for a keyword gap analysis.
Tab 2, Semrush weekly. For each keyword: Overview triggered (Y/N), domain cited (Y/N), citation slot (1 to n), organic position, date. From Position Tracking export or the API. One row per keyword per week.
Tab 3, Search Console weekly. Export the generative AI report by page. Columns: page, AI impressions, date range. One row per page per week.
Tab 4, Rollup. Join Tab 2 to Tab 3 on the target URL from Tab 1. Compute:
Metric | Formula |
Trigger rate | Keywords with Overview / total keywords |
Citation rate | Keywords where you're cited / keywords with Overview |
AI impressions | Sum from Tab 3 for pages in the set |
Rank vs cite gap | Keywords where you rank top 10 but aren't cited, and keywords where you're cited but rank outside top 10 |
WoW change | Each of the above vs prior week |
The rank vs cite gap row is the one worth showing leadership. It's the ranking but not cited problem and its mirror, quantified on your own keywords.
If you want it prettier than a sheet, Looker Studio connects to Search Console natively and takes the Semrush export as a Google Sheets source. But start with the sheet. Most teams never need more.
How do you read the tracker?
Look at three signals. Rising trigger rate with flat citation rate means the surface is growing and you're not on it, which is a structure problem. Rising citation rate with flat AI impressions means you're cited on low-volume terms, which is a targeting problem. Cited-but-not-ranking keywords are your proof that citation is a separate game, and your best case for investing in it. |
The patterns I see most on B2B sites:
High trigger, low citation. Common on comparison and "what is" terms. Seer Interactive's data across 49,000 queries found comparison queries trigger Overviews 95% of the time and question-format queries 86%. If those trigger and you're absent, the fix is on-page. Question-form headings, direct answers, FAQPage schema, and the rest of the AEO checklist.
Cited from outside the top 10. This is the finding that started this whole site's thesis. Pages cited at position 11, 40, and unranked. When you see it in your own data, it reframes the budget conversation. Ranking and citation are different investments.
AI impressions on pages you didn't target. Search Console will surface pages earning AI impressions that aren't in your keyword set. Those are accidental citation winners. Add their keywords to Tab 1 and figure out why they work. Usually it's a well-structured FAQ section or a specific, quotable claim.
Citation slot drift. If you're cited but sliding from slot 1 to slot 4, the sources above you are more specific or more recently updated. Refresh the passage. The query fan-out post covers why passage specificity decides slot.
What can't this tracker see?
Anything outside Google. ChatGPT, Perplexity, Claude, and Bing Copilot citations require prompt-based monitoring, either with a paid tool or a manual cadence. It also can't see AI Overview clicks, because Google doesn't report them, and it can't see the actual prompts behind AI Mode impressions. It's a Google visibility tracker, not a full AI citation tracker. |
For the other engines, the LLM citation tracking post covers what to measure and the AI search visibility tracking post covers the tool options. Bing deserves its own line because it feeds ChatGPT search retrieval, and Bing Webmaster Tools has an AI performance view that most B2B teams have never opened.
Build the Google tracker first. It's free, it's first-party, and it'll tell you within a month whether the AI Overview optimization work is landing. Then decide whether the other engines justify a subscription.
If you want the sheet template with the formulas prebuilt and the Semrush API calls documented, book 30 minutes and I'll walk you through the version I run.
About the author
Harold Bell is Founder and CEO of MQL Magnet, a B2B content marketing agency serving enterprise technology brands. He's the creator of the Engine Optimization Matrix (EOM), author of Partner Over Product, and a Forbes Communications Council member.
Frequently asked questions
What is an AI Overviews tracker?
A system for monitoring which keywords trigger Google AI Overviews, whether your domain is cited inside them, which pages earn AI impressions, and how those change over time.
Does Google Search Console show AI Overview data?
Yes, since June 2026 and globally since August 31, 2026. The generative AI performance report shows impressions from AI Overviews and AI Mode by page, country, device, and date. No clicks or queries.
Does Semrush track AI Overviews?
Yes. AI Overviews are SERP feature code 52. Position Tracking flags keywords that trigger one and shows which domains are cited, with the position indicating the citation slot inside the Overview.
Can I track AI Overviews for free?
Search Console is free and shows AI impressions by page. Keyword-level trigger and citation data requires a rank tracker like Semrush, which most B2B teams already have.
What's the difference between AI Overview position and organic rank in Semrush?
On an AI Overview row, the position is your slot among the cited sources inside the box. Organic rank is your position in the blue links below. They're independent.
Why doesn't Search Console show AI Overview clicks?
Google hasn't exposed click, CTR, or query data for generative AI features and hasn't announced a timeline. The report is impressions only.
How often should I update the tracker?
Weekly. AI Overview composition changes frequently, and weekly cadence catches citation slot drift and new triggers without generating noise.
Can this tracker see ChatGPT or Perplexity citations?
No. It covers Google only. ChatGPT, Perplexity, Claude, and Bing Copilot require prompt-based monitoring with a separate tool or manual process.
What does high trigger rate with low citation rate mean?
The AI Overview surface exists for your keywords but you aren't being selected. It's usually a content structure problem: missing question-form headings, buried answers, or no FAQPage schema.
Should I still track organic rankings?
As context, yes. The most useful view is the gap between ranking and citation, which shows which pages rank but aren't cited and which are cited without ranking.
