Google has been serving complete, normal-looking search results pages with the listings swapped out to the automated tools that SEO professionals use to track rankings, according to a Search Engine Roundtable report by Barry Schwartz published on October 12, 2026 and since updated with statements from Semrush, SE Ranking, SISTRIX and AccuRanker. Nozzle, the rank tracking company whose research anchors the report, counted 8.0% of the pages it collected since September 13 as "poisoned," against 0.17% before.

In Short

Google has started giving fake-looking-real search pages to the robots that check where websites rank, so the tools sometimes report that a site dropped when real searchers would still see it. That matters to every agency, brand and publisher whose reports, dashboards and client invoices are built on those rankings. Some vendors now run extra checks before storing a result, and the numbers they publish suggest the damage depends heavily on how each tool collects its data.

What Schwartz reported

The Search Engine Roundtable article, filed under Google and dated October 12, 2026 at 7:51 am, opens with a blunt assessment. "Google seems to be successfully blocking enough of the attempts of scrapers from trying to scrape data from its search results page over the past few weeks than it has in a long time," Schwartz wrote. He had covered the issue a few weeks earlier and wrote that he was "surprised to see that many tools still struggle with Google Search reporting."

The account rests largely on Derek Perkins, of Nozzle, whom Schwartz described as the most transparent vendor on the subject. According to Perkins, Google is running a three-pronged campaign:

  • IP blocking that is "super wide" and "targets major residential proxy networks"
  • CAPTCHAs used at "a significant increase" in frequency
  • Bad data injection, in which Google places wrong results into suspect pages, "so that when the tools do scrape the results, they get the wrong data in some cases"

The first two are familiar. Rank trackers route their requests through a residential proxy pool so that automated queries resemble ordinary home traffic, and a CAPTCHA is the standard gate a search engine raises when it suspects a machine. The third is new, and it is the one Schwartz singled out. "The third approach, injecting bad data, concerns me the most," he wrote. "I suspect most tools won't pick up on this issue too quickly and thus think all is good when Google may be just toying with them."

Google has made no public statement about any of the three. Schwartz's own reading is that the tools SEO professionals use are not the target. "Again, I don't think Google is taking these efforts to block the tools SEOs use but rather to go after AI bots and SerpApi," he wrote. That is his assumption, labelled as such in the article.

Nozzle's numbers

Perkins published his findings under the title "Poisoned Google SERPs." Schwartz quoted its opening: "Since September 2026, Google has been poisoning the SERPs it serves rank trackers: complete, normal-looking pages whose results have been swapped out. Nothing on the page gives it away. Each keyword's own history does."

The report listed six headline findings: that 8.0% of SERPs came back poisoned; that Google "blocked first, then poisoned"; that "the page looks normal. The results don't"; that social sites and forums fill the gap; that tracked sites fall out of the top 10; and that "it passes within days."

The study itself, as published on Nozzle's website, carries more detail than the Roundtable summary. It is dated October 10, with data through October 9, while Schwartz described it as published early on the morning of October 12. The sample is large. Nozzle collected 27.5 million SERPs between September 13 and October 9, of which 17.8 million had enough history to be judged, across 49 countries and three device types: desktop, iPhone and Android. Counting from August 20, the full sample ran to 63.1 million pages.

The sequence matters. According to Nozzle, requests to Google began failing in large numbers on September 13. The first poisoned pages appeared on September 20, among the requests that still got through. Poisoning peaked between September 27 and 29, at 24.9%, 26.7% and 27.0% of pages, and has run at roughly 5% to 11% a day since October 1.

Nozzle's definition is specific. A keyword's "usual" sites are those that held a top 10 position on at least 60% of its pages between August 20 and September 9. A page counts as poisoned when fewer than half of those usual sites appear anywhere on it, and when they return on a later collection. On poisoned pages, only 20.7% of organic sites were carried over from the previous collection, against 88.5% at baseline.

Who fills the gap

The replacements are not random. Reddit's share of top 10 appearances rose from 20.8% on other pages to 28.8% on poisoned pages, an 8.0 percentage point gain and the largest in the study. Facebook rose from 28.3% to 33.8%, LinkedIn from 6.2% to 9.3%, Instagram from 9.6% to 12.1% and Quora from 5.3% to 6.7%. YouTube, Yelp and MapQuest lost share.

For the companies paying for rank tracking, one figure stands above the rest. A project's own tracked site was missing from the top 10 on 58.6% of poisoned pages, compared with 2.7% at baseline. A site that seemed to disappear for a day in late September and then returned was, in Nozzle's words, most likely a victim of this pattern.

Paid listings behaved differently. On poisoned pages, ads were kept from the previous collection 83.7% of the time, against 70.6% at baseline, according to Nozzle. The organic results were swapped; the ads were not. Nozzle does not offer an explanation, and none of the sources attributes a motive to the difference.

Most incidents were brief. According to Nozzle, 80% of poisoned spells lasted a single collection, and on 77.9% of affected pages the usual sites returned within three days, with a median of two.

Same problem, different rates

The update to the Roundtable article added two vendor posts, and the newer one complicates the 8% figure. AccuRanker published "How AccuRanker Handles Poisoned SERPs" today, written by Peter Emil Tybirk. It describes the same phenomenon and arrives at much lower numbers for its own data.

According to AccuRanker, the large-scale event began late on September 24. On September 25, its sample check flagged 30.2% of desktop pages and 25.7% of mobile pages as poisoned. By September 26 those figures were 4.5% and 4.0%, and between October 1 and 9 both stood at 1.47%. AccuRanker said Google tightened detection again on October 15, mostly on mobile, where the rate reached 2.55% on October 16 before falling to 0.47% on October 17.

The sharpest contrast is a single date. On September 29, when Nozzle measured 27%, AccuRanker put its own rate at 1.4%. AccuRanker's explanation is that the behaviour tracks the way data is collected rather than what is being tracked, meaning that the rate depends on how a vendor's requests look to Google rather than on which keywords or markets a customer follows. Across 22 markets, AccuRanker said, the poisoned share ranged only from 1.1% to 1.7%.

The two vendors also use different tests. AccuRanker's "sandwich check" flags a first page when its top 10 keeps fewer than half the sites seen the day before and fewer than half of those seen the day after. Its figures come from a sample of one in every 200 tracked keywords. Nozzle compares each page against a three-week baseline. The rates are therefore not directly comparable, and neither has been independently audited.

AccuRanker's post also pushes the timeline back. It reports smaller rounds of the same pattern in its stored data from late January 2026, and levels between April and July running at about seven to ten times their 2025 level, when whole-page flips occurred roughly once per 1,000 daily checks or fewer.

What the vendors told customers

Schwartz wrote that when he first reported the problem, most tools he contacted did not respond before publication. Several later posted notices. According to the article, SISTRIX commented, DataForSEO posted on its status page, Advanced Web Rankings and Morningscore published notes, and SE Ranking said for Google rankings it was "missing positions for some keywords" but that most of the issue had been resolved. Schwartz also wrote that some vendors "confused the goto redirects with the new issues occurring," when "the goto redirects are the easy part for the tools to handle."

SE Ranking followed up on October 16 with a post on X. "Google has recently been returning incomplete or unreliable results to rank trackers across the industry," the company wrote. "This can cause keywords to show false drops or appear as not ranking, even when their actual positions haven't" changed.

Semrush sent Schwartz a statement:

"We can confirm we've been seeing the "poisoned SERPs" issue with Google returning inaccurate results since late summer. As soon as we noticed it, our team put additional checks in place to filter and double-check the results we return, and we keep adjusting and extending them as Google's behavior evolves."

The statement continued: "The impact on the rankings our users track has been limited, and we have seen no interruptions to the service. There can still be individual cases, most often beyond the first page of results, and we encourage users to report any inconsistencies to our support team so we can look into each one."

"Since late summer" places Semrush's first observation earlier than Nozzle's September 20 date, and closer to AccuRanker's account of a slow build through 2026.

SISTRIX narrows its scope

SISTRIX's post, "Why Google ranking data collection is changing," was linked in the Roundtable update. It is dated October 7, 2026 and signed by Johannes Beus. According to SISTRIX, Google has returned results for some queries since September 2026 that look genuine but list different websites. The company said it publishes only results that pass several quality checks, recollects keywords that fail until a verified result exists, and has disabled on-demand updates since September because they would bypass those checks.

SISTRIX also said it would concentrate on the top 10, because, according to the company, positions beyond that can no longer be compared reliably over time. It described top 10 rankings and its Visibility Index as still reliable, and attributed Google's changes to an effort to limit new competitors such as ChatGPT and to stop search results from being used to train open-weight models, adding that analysis tools are caught up in it rather than targeted. That is SISTRIX's interpretation. Google has not confirmed it.

What practitioners saw

The Roundtable article collected posts from SEO practitioners on X and Bluesky, and they describe a period in which the tools disagreed with each other and with reality.

On October 8, Lily Ray replied to Google Search Central's notice that a rollout "was complete as of October 8, 2026" with: "Trying to share some findings on this but rank trackers are too unreliable at the moment." The rollout in question was the September 2026 spam update, which began on September 24 and ended on October 8. Its start coincides exactly with AccuRanker's onset date, which means that any ranking drop recorded in those two weeks could have three causes: a genuine spam demotion, a blocked collection, or a poisoned one.

Glenn Gabe illustrated the problem on October 7 with a site he said "got hammered on 9/29 during the spam update," a drop "100% confirmed by the site owner." According to Gabe, "ahrefs accurately picked up the drop, Semrush way late picking it" up. Ryan Jones wrote on October 1: "we're getting to the point where nobody will have rank tracking anymore because of the crazy amount of AI scraping going on."

Replies to Schwartz's October 9 call for experiences ranged from "They don't work at all atm" (BowTiedWookie) and "Nothing matches the SERPs" (Marty Eigner) to more measured accounts. Sarah Blocksidge wrote that at the end of September and early October "they were definitely broken," that "some stuff seems to be coming back ok (in SE Ranking and Ahrefs)," and that she "never really saw issues with Semrush." Daniel Skov Jacobsen summarised the economics: "Looks like trackers are paying for Google making SERP scraping expensive: num=100 gone, JS required. Most now sample instead of counting, so some daily swings are measurement noise."

One post concerned an industry fixture. On October 9, Dr. Pete Meyers wrote on Bluesky: "I have documented and handed off MozCast as best as I'm able, so I hope it will continue. Google certainly doesn't make it easy to maintain." MozCast, the daily Google volatility index Meyers built at Moz, has long served as a public reference for ranking turbulence. The post does not say who now runs it.

Can the industry survive?

The question has been asked before. Each previous turn of the screw raised the cost of collecting Google results without ending the business. Google began requiring JavaScript for search in January 2025. It dropped the num=100 parameter on September 14, 2025, so that collecting 100 results took ten requests instead of one. Ahrefs needed about six weeks to restore top 100 tracking afterwards, and DataForSEO's workaround, which it said cut costs by 80%, was blocked within five days. Then came the goto redirect links, which Perkins said turned a single five-page ranking into 500 to 1,000 requests if every link is resolved. SISTRIX, which says the redirects date to June 2026, absorbed them through one extra server request per URL.

Those were cost problems. Vendors could pay more, sample more, or collect fewer pages. Poisoning is a different category of problem, because it attacks the credibility of the output rather than its price. A blocked request produces an empty cell that a vendor can see. A poisoned page produces a confident number. Nozzle's study makes the distinction directly: a refused request is a gap, while a poisoned page is stored by the tool as if it were real.

The evidence in the sources points in two directions at once.

On one side, the vendors that have published detail say they can detect and filter most of it. AccuRanker reports rates below 1% after mid-October validation. Semrush says the effect on its users has been "limited." Swapped pages are also short-lived by every vendor's measurement, and a second fetch returns normal results about three times in four, according to AccuRanker. The goto redirects were handled within days by most vendors, and the num=100 change within weeks.

On the other side, every one of those defences costs more requests, more checks and less frequent data. SISTRIX has retreated to the top 10. AccuRanker says one in 20 desktop pages and one in 12 mobile pages lack enough history for its comparison to work at all. Jacobsen's observation that most tools "now sample instead of counting" points to the same trade. A practice built on daily, keyword-level, full-depth measurement is moving toward first-page sampling with error bars. And the most widely cited public volatility index has just changed hands.

PPC Land's own reading of the vendor figures, which is a calculation and not a measurement, is that the gap between Nozzle's 27% and AccuRanker's 1.4% on September 29 is itself the most important number in the story. Two companies collecting the same search engine on the same day saw rates that differ by a factor of about 19. Unless vendors publish their error and sampling rates, customers cannot tell which of those worlds their own data came from. AccuRanker has now published daily rates for its sample. DemandSphere did something similar earlier this month, after it reported wrong results on about 11% of its daily checks and added an observability layer to its platform.

The first-party alternative

Schwartz pointed to the obvious substitute. "Google does offer a whole portal of information about rankings and more in Search Console," he wrote. "It might not be enough for what most SEOs need but Google does give us a lot of data there. Well, not click or query data for generative AI search, like AI Overviews and AI Mode..."

That limit is documented. The generative AI report Google added to Search Console for all sites on August 31, 2026 shows impressions without clicks or queries for AI Overviews and AI Mode. Search Console also reports only on sites the user owns or has verified. It cannot show a competitor's position, a market's share of voice, or which forum thread displaced a brand on a given morning. Those questions have only ever been answerable through third-party collection, which is the very thing Google is now degrading.

That is the structural tension. Google has been fighting automated access to its pages in court as well as in code. It sued SerpApi in December 2025, alleging that the company evaded the SearchGuard bot detection system, and when the first complaint was dismissed in July 2026 it filed an amended one; SerpApi moved to dismiss again in August. PPC Land noted at the time of the original filing that Google is itself one of the largest collectors of other people's content. Whether Google's target is AI companies, SerpApi or bulk web scraping in general, SEO tools sit on the same infrastructure, and nothing in the sources suggests Google distinguishes between them.

Why this matters for marketers

For agencies, the immediate exposure is in reporting. A client deck built on September and October 2026 ranking data may contain drops that never happened in front of real searchers, and the period overlaps a confirmed spam update that produced real ones. Nozzle's figures suggest the false drops are concentrated exactly where clients look first: a tracked site falling out of the top 10 for a day or two while Reddit and Facebook take its place.

For advertisers, the data is narrower but still relevant. Competitive intelligence on paid listings, organic share of voice reporting and AI Overview monitoring all draw on the same scraped pages. Nozzle's observation that ads were kept more consistently than organic listings on poisoned pages is one data point and has no stated cause.

For the SEO tool market, the question of whether the industry can survive does not have a sourced answer. The sources support a narrower conclusion: the vendors that have published methodology say they can detect most swapped pages, at rising cost and with reduced depth, and Google has not said whether it intends to continue, expand or stop. AccuRanker's post says plainly that Google has made no statement on how long this will last.

Timeline

  • January 2025 - Google begins requiring JavaScript for search, disrupting several rank checking tools.
  • September 14, 2025 - Google drops the num=100 parameter, forcing ten requests where one had returned 100 results.
  • October 29, 2025 - Ahrefs begins restoring top 100 tracking, about six weeks after the num=100 change.
  • December 19, 2025 - Google sues SerpApi, alleging circumvention of SearchGuard.
  • Late January 2026 - AccuRanker later identifies smaller rounds of swapped pages in its stored data.
  • April to July 2026 - Swapped-page levels in AccuRanker's data run at about seven to ten times their 2025 level.
  • June 2026 - Goto redirects in Google results begin, according to SISTRIX.
  • July 20, 2026 - Court dismisses Google's original complaint against SerpApi.
  • August 24, 2026 - SerpApi moves to dismiss Google's amended complaint.
  • August 26, 2026 - Google confirms the goto redirect rollout; Nozzle measures 500 to 1,000 requests per five-page ranking.
  • August 31, 2026 - Search Console's generative AI report reaches all sites, with impressions but no clicks or queries.
  • September 13, 2026 - Requests to Google begin failing in large numbers, according to Nozzle.
  • September 20, 2026 - First poisoned pages appear in Nozzle's data.
  • September 24, 2026 - Google's September spam update begins; AccuRanker records the onset of large-scale swapped pages late that day.
  • September 25, 2026 - AccuRanker flags 30.2% of desktop and 25.7% of mobile sample pages.
  • September 27 to 29, 2026 - Poisoning peaks at Nozzle at 24.9%, 26.7% and 27.0%; AccuRanker measures 1.4% on September 29.
  • October 1, 2026 - Ryan Jones posts that "nobody will have rank tracking anymore."
  • October 7, 2026 - SISTRIX publishes "Why Google ranking data collection is changing"; Glenn Gabe posts on tool discrepancies.
  • October 8, 2026 - Google completes the spam update; Lily Ray calls rank trackers "too unreliable."
  • October 9, 2026 - Dr. Pete Meyers says he has handed off MozCast; Schwartz asks practitioners for their experiences.
  • October 10, 2026 - Nozzle's "Poisoned Google SERPs" study is dated, with data through October 9.
  • October 12, 2026 - Search Engine Roundtable publishes Schwartz's report.
  • October 13, 2026 - PPC Land reports DemandSphere's 11% spurious-results figure.
  • October 15, 2026 - Google tightens detection again, mostly on mobile, according to AccuRanker.
  • October 16, 2026 - SE Ranking tells customers Google is returning "incomplete or unreliable results."
  • October 19, 2026 - AccuRanker publishes "How AccuRanker Handles Poisoned SERPs"; Schwartz's article is updated with vendor posts and a Semrush statement.

Summary

Who: Google; rank tracking vendors Nozzle, Semrush, SE Ranking, SISTRIX, AccuRanker, DataForSEO, Advanced Web Rankings and Morningscore; Search Engine Roundtable's Barry Schwartz; and SEO practitioners including Lily Ray, Glenn Gabe, Ryan Jones and Dr. Pete Meyers.

What: Google has been serving rank trackers complete-looking search pages with swapped results. Nozzle counted 8.0% of its pages as poisoned since September 13, peaking at 27.0% on September 29, with tracked sites missing from the top 10 on 58.6% of those pages. AccuRanker measured far lower rates with its own collection, and Semrush and SE Ranking confirmed the issue and said they filter results.

When: Blocking intensified from September 13, 2026 and poisoned pages appeared from September 20 in Nozzle's data and September 24 in AccuRanker's. Schwartz reported on October 12, 2026, and the article was updated with vendor posts through October 19.

Where: Google Search, across 49 countries and desktop and mobile devices in Nozzle's sample and 22 markets in AccuRanker's.

Why: Google has not explained its actions. Schwartz and SISTRIX believe the target is AI companies and SerpApi rather than SEO tools. For marketers, the consequence is that ranking reports from late September and October 2026 may show drops that real searchers never saw, and the scale depends on how each vendor collects and checks its data.