Google AI Overviews How to Track Them and Win Placement

The shift

Why AI Overviews change the way visibility is measured

A page can hold a strong position in the organic listings and still not be one of the sources linked inside the summary shown above them. Position alone therefore no longer describes what a searcher saw.

The click data explains why this matters. A Pew Research Center study of 900 US adults' browsing in March 2025 found that users clicked a traditional result in 8% of visits to a results page with an AI summary, against 15% on pages without one. A link inside the AI summary was clicked in 1% of visits, and the browsing session ended on that page more often when a summary appeared (26% of visits, against 16%).

The same study shows which queries are affected. An AI summary appeared on about 18% of the 68,879 searches analysed. It appeared on 8% of searches of one or two words and on 53% of searches of ten or more words, and on 60% of searches that began with who, what, when or why. Tracking should therefore be segmented by query type rather than treating every keyword the same way. These figures describe US desktop-and-mobile browsing in one month and should not be read as rates for any particular site or country.

For wider context on measuring citation-based visibility across several AI platforms, see the guide on how to measure AI visibility.

The metrics

What an AI Overviews tracker actually monitors

A tracker built for this purpose checks a different set of outcomes from a normal rank tracker:

• Whether an AI Overview appears at all for a given query.
• Whether the tracked domain is one of the sources linked inside it, which is the core of AI citation tracking.
• Which competitor domains are linked on the same queries.
• The wording of the generated answer, where the tool stores it, so changes can be compared between runs.

Citation position needs careful reading. Google's Search Console documentation states that an AI Overview occupies a single position on the results page and that all links inside it share that position. Two competitors cited in the same overview therefore sit side by side rather than in a strict first-and-second order, and a "position in the citation list" figure from a third-party tool is a display order, not a ranking.

Results also vary between runs. A SparkToro and Gumshoe study of 2,961 prompts across ChatGPT, Claude and Google's AI features, reported by Search Engine Journal, found that the same list of brands came back in fewer than 1 in 100 repeated runs for Google AI and ChatGPT, and the same order in fewer than 1 in 1,000. How often a brand appears across many runs was more consistent than where it appears in any one run. A single check is a snapshot; a visibility percentage over repeated runs is the more dependable measure. The related metric is covered in the guide to AI share of voice.

Process

How to track AI Overview appearances for target keywords

Reliable tracking combines several checks rather than a single dashboard.

1. Start with Search Console. Google's Generative AI performance report shows impressions for a site's pages in AI Overviews and AI Mode, with Pages, Countries, Dates and Devices dimensions. It has no query dimension, so it shows how much exposure exists and where, but not which prompts or competitors sit behind it. Google's help page lists the report's remaining limits and says it reached all sites by 31 August 2026.

2. Add prompt-level checks for the gap. Search Console cannot show which queries trigger an overview or which competitors are cited next to a site. That is the job of manual checks or a dedicated tool: run a fixed list of keywords on a schedule, record whether an overview appeared, whether the domain was cited, and which other domains were. Google's own guidance warns site owners to be wary of third-party tools claiming to report Google's internal metrics, so tool output should be treated as observed results from sample runs rather than Google data.

3. Control the conditions. Results differ by location, device and signed-in state. Searching while logged in to a personal account, or through a VPN outside the target market, can return a different page from the one a customer sees. Run checks from a clean session set to the target city and keep device (mobile or desktop) consistent, or record each separately. The Search Console report's Devices dimension can be used to see whether the split matters for a given site.

Signals worth recording consistently, rather than ad hoc:

• Whether an AI Overview appears for each target query, on a fixed schedule.
• Whether the domain is cited, and which competitor domains are cited alongside it.
• The share of repeated runs in which the domain appears, rather than a single result.
• Query length and type, since Pew found overview frequency differs sharply by both.

Workflow choice

Choosing between Search Console, manual checks and a dedicated tool

Manual checks work for a handful of priority keywords but do not scale: checking even fifty keywords by hand, on two devices and repeated several times, quickly becomes a reporting burden. A dedicated tool runs the same checks automatically and stores history, so trends become visible rather than anecdotal. Search Console remains the only first-party source, but it reports totals, not prompts.

Terminology

Google AI Mode and AI Overviews: what's the difference

Google describes both under its AI features documentation, including how pages become eligible to appear as supporting links: a page must be indexed and eligible to be shown with a snippet, with no additional technical requirements. The guidance on AI features for site owners also notes that these features may use a "query fan-out" technique, issuing several related searches across subtopics to assemble a response.

The term "SGE" (Search Generative Experience) still appears in some tool names. Google's May 2024 announcement says AI Overviews began as an experiment in Search Labs before launching in the US, so tools marketed as "SGE tracking" and tools tracking AI Overviews are, in practice, looking at the same feature. A tool advertising AI Mode monitoring should state whether it records AI Mode responses or only AI Overviews.

Optimisation

What helps a page get featured in an AI Overview

No formula guarantees inclusion, and any claim promising one should be treated with scepticism. Google's guide to optimising for generative AI features is the primary source, and its position is that the practices are the same as for Search generally. The points it makes that matter here:

• A page must be indexed and eligible to appear with a snippet to be a supporting link.
• No special markup, AI-specific text file such as llms.txt, or content "chunking" is required.
• Content that is distinctive and not a commodity is favoured over content that restates what is already widely available.
• Mass-producing pages for every fan-out variant of a query risks being treated as scaled content abuse.
• Structured data, where used, must match the visible text on the page.

Measurement and optimisation connect through the citation gap in the example above: topics where an overview appears but the site is not cited are the ones to review. The guides on why AI cites some businesses and ignores others and how to improve AI answer inclusion cover that step.

Quality control

Common measurement mistakes that skew data

Comparing data without controlling location or device

Comparing AI Overview data across locations or devices without controlling for either produces numbers that look inconsistent for no real reason.

Treating a single run as the result

Repeated runs of the same prompt can return different sources, as the SparkToro and Gumshoe study showed. A single check is a snapshot, so conclusions should rest on the share of repeated runs.

Treating a citation as a click

A citation inside an overview is not a visit. Pew found a link inside an AI summary was clicked in 1% of visits to pages that showed one, so citations should be read alongside Search Console clicks and analytics rather than assumed to drive traffic.

Ignoring query fan-out

Google's documentation notes that a single query can lead to several related searches behind the scenes, so one tracked keyword may not reflect everything the response draws on. Tracking a set of related phrasings gives a fuller picture than one keyword.

Sources

Primary references

• Google Search CentralAI features and your website
• Google Search CentralOptimizing your website for generative AI features
• Search Console HelpGenerative AI performance report
• Search Console HelpPosition in Search Console data
• GoogleGenerative AI in Google Search (May 2024)
• Pew Research CenterGoogle users are less likely to click on links when an AI summary appears
• Search Engine JournalAI recommendations change with nearly every query (SparkToro)
• SEO Analyser BlogAI citation tracking guide