Semrush AI Overview rank tracking, compared with a second measurement

Semrush's knowledge base says Position Tracking shows which SERP features you appear in, including AI Overviews. You create a campaign inside a project with your website URL, a location, a device type and your keywords, then open the Overview tab and use the SERP features filter to show keywords that triggered an AI Overview where your domain appears as a source. Its AI Overview guide says Position Tracking updates daily.

Semrush also describes Organic Research, Domain Overview, Keyword Overview, the Keyword Magic Tool and Semrush Sensor as places to see AI Overview data. The Semrush MCP connector lists a position_tracking tool, but this page does not assume which of those AI Overview filters the connector returns, so ask Claude to report exactly what it received.

Semrush's knowledge base notes that when your domain is cited as a source in an AI Overview, Position Tracking records it as a number one ranking, whatever your page ranks in the organic results below. It adds a SERP features toggle to view rankings with or without AI Overviews counted. That is a defined convention, and it means the Semrush position can differ from the organic rank another tool reports.

Linkeddit keeps the two apart. In the test run, cookieless analytics has a rank of null (not in the Google top 10) while the AI Overview shown for it cited plausible.io. A tool that counts the citation as a top position and a tool that reports organic rank will tell different stories about the same keyword, and the table should show both columns.

search_visibility_get returns, for each tracked keyword, the latest rank, the change since the previous check, whether Google showed an AI Overview, whether it cited the project and the top domain. On the test project 25 keywords are tracked, and three of the four shown had an AI Overview that cited the project.

view=history gives the rank series over up to 52 weeks. view=aio_impact splits keywords into those with and without an AI Overview and compares Search Console clicks before and after, but it needs Search Console connected. Semrush's own guide notes that Search Console folds AI Overview clicks and impressions into overall performance, which is why a separate record matters.

Treat each reading as a sample. Linkeddit states in its results that a rank of null is a result, not a failure, and that its checks cover the keywords you track, not the whole web. Semrush's guide says an AI Overview appearing on a keyword you rank for does not mean your site is featured in it, which is why it separates the two filters.

Semrush also sells an AI Visibility Toolkit for brand mentions in AI platforms, and Linkeddit measures ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Use this report for Google rank and the Overview, and use the competitor and citation workflows for the engine answers.

Some SEOs build the check by hand. "Right now I’m checking a fixed list of buying questions every month, noting whether an AI Overview appears and who it cites, and pairing that with branded search and GA4 referrals from AI platforms," described an ecommerce SEO on Reddit. That routine shows why a second, repeatable measurement helps when you compare it against a tool's reading.

Semrush's AI Overview guide says Google Search Console rolls AI Overview clicks and impressions into overall performance, so there is no easy way to pull them out, and reports that AI Overviews show for over 50% of queries in Semrush Sensor. Neither fact tells you what happened to your own clicks.

Linkeddit's aio_impact view of search_visibility_get compares Search Console clicks before and after for keywords with and without an AI Overview. It defaults to 28 days each side, accepts up to 180, and returns both groups, a verdict with its version and the caveats. It needs Search Console connected, so treat it as a comparison to read, not proof of cause.

Run the report weekly and keep the dated tables. A keyword that gains an AI Overview in one week and loses clicks in the next is worth a closer look, and a saved sequence of readings from both tools shows it far better than a single snapshot. Ask Claude to note the check date beside every number.