Semrush says Google is serving poisoned results to rank trackers

Semrush has said that Google is sometimes answering automated search requests with results pages that look normal but do not match what a person typing the same query would see, and that the issue is industry-wide, affecting SEO data providers beyond its own tools. The statement, dated 27 October 2026, describes what Semrush calls poisoned SERPs and sets out how the company is filtering them out of its database.

Semrush says the issue is industry-wide and affects data providers across the SEO space, not only its own tools. Google has not commented publicly. Industry discussion points to Google limiting large-scale data collection for AI systems as the likely cause, but that is an inference, and nothing from Google confirms it.

Two poisoned shapes, and why the second corrupts quietly

According to Semrush, a poisoned results page takes one of two shapes. In the first, the entire page is plainly wrong for the search, and the mismatch is visible on page one. In the second, the opening results look right, and the irrelevant pages only start appearing lower down, with a weak connection to what was searched.

The second kind is the one that corrupts a rank report quietly. A tracker reading that page records a real-looking position for a page that a human searcher would never have been shown.

Semrush's own example is a search in Brazil for a specific model of baby stroller, which returned a dental supply store's equipment page. The two share the Portuguese word "carrinho". A keyword report built on that page would show the dental store ranking for a pram query it has nothing to do with.

Invalid excluded, doubtful collected again

Semrush says results that are plainly invalid are caught while the data is being gathered and never reach its database. It has also adjusted its data sources based on data quality, and now runs extra checks that mark doubtful rankings for a second collection.

The company says it will keep refining those processes as Google's behaviour changes, and that the statement reflects its understanding as of 27 October and may be updated. It also says it will keep investing in both traditional ranking data and AI answer tracking, covering Google results, AI Overviews and AI platforms such as ChatGPT.

For a business owner, the practical reading is that some of the odd movements in recent ranking reports may be artefacts of collection rather than real changes in Google, and the vendor itself is saying so.

9.5 out of 10 on Semrush's sensor, from the same automated collection

The timing matters. Rank tracking tools recorded a spike in ranking volatility on 27 October, along with a rise in chatter from site owners about a possible Google update. Google has not confirmed any update. This followed earlier unconfirmed spikes on 24 October, 20 and 21 October and 16 October, and the September 2026 spam update, which Google ran from 24 September to 8 October.

Semrush's own volatility sensor, which scores on a 10-point scale, read about 9.5 on 27 October, up from 7 the day before. One site owner described large keyword gains without matching traffic gains, and another high traffic volume with conversions well down.

Put the two stories together and the caution is obvious. The sensors that say Google is volatile are built from the same kind of automated results collection that Semrush says is sometimes poisoned. A spike could be a Google update, a change in what Google serves to bots, or a mix of both, and the tools cannot easily tell those apart. The same pattern appeared with the missing AI Overview label on some desktop results, which could have shown up as a two-day dip in AI visibility reports and deserved a manual check before anyone acted on it.

Check unrelated keyword moves in Search Console's Performance report

Semrush's advice to its customers is specific: "If you see unexpected ranking changes, especially for keywords unrelated to your pages, compare against your own data, such as Google Search Console." It also asks customers to send its support team the affected domain and keywords so the data can be reviewed.

That is sound advice for anyone reading an SEO report, whichever tool produced it. Search Console's Performance report records the queries Google actually showed your pages for, the impressions and clicks that followed, and the average position, all from Google's side rather than from a scraper. It cannot be poisoned in the way a scraped results page can.

Search Console has its own limits. Its data arrives with a lag, and straight after the September spam update finished, the Performance report fell more than 64 hours behind before catching up, the subject of our post on Search Console data delays. Google's John Mueller also said this week, in a public reply to a site owner whose Page Indexing report had not updated since 16 October, that indexing does not change quickly and that when it does, the change shows up in the Performance report first. He added that the dates on each report are there so users can see how fresh the data is, and that Google prioritises faster updates where it thinks they are actually useful.

So the hierarchy for a business owner is: Search Console for what Google actually showed, with a known lag; the rank tracker for a faster but noisier signal; and enquiries, calls and sales as the measure that settles any argument between the two.

Whatever the dashboard shows, a drop leaving enquiries untouched is not real yet

If a ranking dashboard shows a sudden move, this is the order we suggest a business owner, or the person reading reports on their behalf, works through:

• Look at which keywords moved. If pages are suddenly gaining or losing rank for terms that have nothing to do with the business, treat the data as suspect first and the site second.
• Open Search Console's Performance report for the same date range and the same queries. If impressions, clicks and average position are steady there, the tracker is the thing that changed.
• Check real outcomes: enquiries, calls, form fills and sales in your analytics or CRM. A ranking drop that leaves leads untouched is either not real yet or on terms that do not matter.
• Ask whoever sends your reports which tool collected the ranking data and what that vendor has said about data quality since 27 October.
• Send the affected domain and keywords to the tool vendor so the vendor can review the data, as Semrush has asked its customers to do.

The same discipline applies to AI visibility reports. Semrush says it is investing in tracking brand presence across AI Overviews and AI assistants. That data is also gathered through automated requests. The Semrush AI Overviews study published earlier this month found AI Overviews on 88% of tracked commercial keywords in US desktop data, which makes those reports worth watching, and worth checking the same way.

Semrush's statement leaves Australian businesses paying for SEO three fair questions

Many Australian businesses paying for SEO receive a monthly report with a ranking table near the top. After this statement, three questions are fair to ask:

• Which tool supplies the rankings, and has the agency reviewed its data since Semrush's statement?
• Does the report put Search Console impressions and clicks alongside the tracker positions, so the two can be compared?
• Are leads and revenue reported with the same prominence as positions?

Rank tracking is still useful. It shows competitors, it shows the shape of a results page, and it gives an early signal days before Search Console data lands. What has changed is that a position in a scraped report now carries a data-quality risk the vendor has acknowledged, so it should sit beside first-party data rather than above it.

The practical next step: pull the Search Console Performance report for the last 28 days, compare it against the ranking table in your most recent SEO report, and list every keyword where the two disagree. That list is what to send to your agency or tool vendor.