Google has a new weapon against the bots that scrape its search results: fake results. Since mid-September, Google has deliberately served altered search engine results pages to automated scrapers, a tactic rank-tracking firm Nozzle dubbed “SERP poisoning.” The goal was a surgical strike against competing language models and AI visibility tools. But according to a Search Engine Land analysis by Ryan Jones, Senior Vice President of SEO at Razorfish, published on November 3, the poison has breached containment. To get clean data, scrapers now search the same keyword three to five times — and the flood of repeat searches may be contaminating the Google Search Console metrics that website owners rely on to measure success.
Instead of blocking bots with CAPTCHAs or IP bans, Google feeds automated scrapers results pages that look legitimate but have the actual ranking websites swapped out for irrelevant but plausible links, along with large amounts of user-generated content such as forum threads and social media profiles. Research from Nozzle, which analyzed 27.5 million search results pages, found that roughly 8% of SERPs were deliberately poisoned with misleading data. Ahrefs reported that 5% to 20% of results sent to data providers were affected, depending on the provider, and major SEO platforms including Semrush, Ahrefs and BrightLocal have confirmed that the altered results are significantly affecting their accuracy.
Poisoned pages also contain significantly fewer AI Overviews than normal results, which the article suggests is likely intentional, given how many tools specifically scrape Google to track that feature. Google has not publicly confirmed that it is intentionally serving fake results, or explained its motives; the poisoning explanation is the interpretation of Nozzle, the affected platforms and the article's author, based on the pattern of the altered pages. What is firmly established is that automated Google result collection has become unreliable since mid-September, and that the providers know it.
Scrapers have adapted with brute force. Poisoning is usually temporary: in 80% of observed cases it lasts for just one search request, and in only 15% of cases does it persist for two or more consecutive requests. Because poisoned results are randomized and highly volatile, tools have adopted consensus scraping — repeatedly searching for the exact same keyword until they get two nearly identical or matching sets of results. Google is unlikely to serve the same randomly generated fake SERP twice in a row, so a match means the tool has found the real rankings. The irony, as the article puts it, is that Google's attempt to stop scrapers is forcing them to scrape three to five times as much data.
That workaround creates a second problem inside Google Search Console. Google likely excludes poisoned results from Search Console, but the repeated searches themselves are the issue: some scrapes evade Google's bot detection and return real results, and those searches can then appear in website owners' Search Console data. If your site ranks on page one for a keyword, every additional scrape can generate a fake impression, so a keyword checked once a day might now be checked three times. And because bots do not click, the inflated impressions come with no clicks, so click-through rates fall — and website owners may wrongly conclude their titles and meta descriptions have stopped working. Because Google uses click data in its Navboost ranking system, undetected bot activity could even feed back into the algorithm itself.
The immediate lesson is to stop trusting single-day rank movements. Most poisoned results snap back within three days, so a one-day plunge that recovers is almost certainly noise, not a ranking loss. Check Search Console first, since it is Google's own data about real searches and real clicks, then look at the three-to-seven-day trend in your rank tracker, and only then consider whether a page actually needs work. Rewriting a page because a poisoned SERP said it dropped is the most expensive mistake this episode can cause.
The deeper question is whether real users are affected too. The article runs the math: Google processes roughly 8.5 billion searches a day, so even if its bot detection is 99.9% accurate, the 0.1% error rate means 8.5 million legitimate searches a day could receive poisoned results; at 99.99% accuracy, the figure is still 850,000. Until Google offers a public, paid Search API — something it has little incentive to do — the scraping war will continue, and the data SEOs rely on will stay noisy. The honest response is to treat rank data as a trend, Search Console as the ground truth, and business results as the final judge.
Sources: Search Engine Land, November 3, 2026.
Related reading on this site: Google Organic Click Share Hits a 15-Month High as Zero-Click Searches Fall, Reddit Begins Testing Ads in Its Own Search Results, 83% of ChatGPT Shopping Picks Also Rank in Google's Top 40, Semrush Finds, Google Rewrites Its 'Crawled – Currently Not Indexed' Guidance, Search Console's Device Filter Is Now Multi-Select, and Email and Paid Search Are Growing as Organic Search Shrinks.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial