I know, and you know that rank trackers and software tools are shaking in their boots because it appears as if Google is messing with the crib. Like you, I often find myself looking at rankings through various tools, and if you have noticed, like I have, that rankings have looked drunk since mid-September, part of that is real, and part of it is your rank tracker being fed garbage.
We've had four unconfirmed Google updates in about two weeks (Oct 16, Oct 20-21, Oct 24, and a bigger spike overnight Oct 26-27, per Search Engine Roundtable). Google hasn't confirmed any of them. Meanwhile, on Oct 27, Semrush became the latest SEO platform to confirm that Google has been serving "poisoned" results to automated rank trackers.
So here's the uncomfortable question for anyone who buys, sells, or reports on links: how many link decisions got made this fall off a chart that was partly fiction?
What actually happened
Google appears to be feeding scrapers bad results, and every major rank tracker now says so. Search Engine Land reported on Oct 27 that Semrush calls the problem industry-wide. Ahrefs, AccuRanker, BrightLocal and others have reported it since September.
Semrush describes two flavors. Some results pages are flat-out irrelevant. Others look normal up top and go sideways further down. My favorite example: a Brazilian search for a specific baby stroller returned a dental supply store, because both products share the Portuguese word "carrinho." Close enough, Google.
The biggest dataset comes from Nozzle, which checked 56.1 million Google results pages across 53 countries from Aug 20 to Oct 21:
- 22.6% of results pages checked on both page 1 and 2 were flagged as poisoned since Sept 13, up from 0.28% in the baseline period.
- The worst single day was Sept 30, at 43.5%.
- Page 2 got hit harder (17.9%) than page 1 (12.2%). But page 1 was not safe.
- 67.9% of affected results pages went back to normal within three days.
- Facebook, Instagram, YouTube and Quora showed up more often in the fake results than in the real ones.
Other trackers report smaller, but very real, numbers:
| Provider |
What they reported |
| Ahrefs |
5% to 20% of results affected, depending on data provider (Oct 22) |
| AccuRanker |
40% of first desktop checks poisoned on Oct 24; 3.2% after filtering, under 1% the next day |
| SEOmonitor |
About 8% of daily results inaccurate or incomplete |
| DemandSphere |
6.7% of collected results pages flagged on Oct 23 |
| BrightLocal |
Mostly rankings below position 10 |
These come from different samples and methods, so don't line them up and average them. Google has not commented, and hasn't confirmed it's doing this on purpose. The popular theory is that it's an anti-scraping move aimed at AI data collection, and SEO tools are collateral damage.
One more caveat in fairness: Search Engine Land is owned by Semrush. The story checks out against everyone else's data, but it's Semrush's site covering Semrush's admission.
Why link builders should care
A lot of link programs are judged by a rank chart, and that chart has had a fake-data problem for six weeks.
Think about how this plays out. A client logs in Monday, sees 30 keywords fall off a cliff, and asks what their link vendor is doing. The vendor scrambles. Maybe someone buys a batch of quick links to "stop the bleeding." Maybe someone pauses a campaign that was working fine. Then, two-thirds of the time, the results pages snap back on their own by Thursday.
That's not a strategy. That's reacting to weather.
And it cuts both ways. Fake gains are just as dangerous as fake drops. One site owner in the SER comments reported huge keyword gains without matching traffic gains. If your vendor is claiming credit for a spike that showed up the same week trackers were broken, ask for a second source.
The number that held still
While rankings twitched and trackers got poisoned, an editorial link on a real publication stayed exactly where it was.
Even the backlink tools had a messy week. On Oct 26, Majestic announced it had pulled hundreds of billions of backlink URLs out of its index. The culprit was a network of 10,000+ spam sites built to stuff link-building ads into backlink reports. Majestic also retuned Trust Flow, and some sites saw big swings in their numbers. Your metrics moved. A link from a site people actually read did not.
That's the point I keep coming back to. A ranking is a reading. A link is an asset. Readings get noisy. Assets stay on the books.
It also matters more now, because links pay in two currencies. They still drive rankings. And the same earned placements are what AI answers cite: Muck Rack's latest data (Oct 28) puts earned and third-party sources at 84%+ of AI citations. Nobody's rank tracker measures that second currency well yet. Another reason not to judge a link program by one chart.
What to do instead
You don't need to throw out your rank tracker. You need to stop letting it make decisions alone.
- Wait three days before reacting. Nozzle found 67.9% of poisoned results pages bounced back within three days. A drop that's still there next week is worth a conversation. A drop that showed up this morning is worth a coffee.
- Check Search Console first. Every provider, Semrush included, is telling customers to compare against GSC. Impressions and clicks come from Google itself, not a scraper.
- Look at the keywords, not just the count. Poisoned results tend to show up for unrelated junk (dental supplies for strollers). If the "new competitor" outranking you is a random Quora thread, be suspicious.
- Report links as links. Referring domains earned, the publications they're on, and the traffic those pages send. Those numbers don't get scrambled by Google's anti-scraping fights.
- Be skeptical of any vendor claiming a big win or blaming a big loss this fall. Including us. Ask what else confirms it.
The bottom line
Google's fight with scrapers turned rank tracking into a weather report this fall. Fine. Check the forecast, but don't rebuild the house every time it rains.
The links you earned on real publications are still there, still passing value in rankings, and still showing up where AI goes looking for sources. That's the number worth watching.
Sources
- Semrush confirms 'poisoned' Google results are impacting SEO tools, Search Engine Land, Oct 27, 2026
- Nozzle: poisoned Google SERPs research (figures as reported by Search Engine Land)
- Google search ranking volatility spike, Search Engine Roundtable, Oct 27, 2026
- Fighting Ref Domain Spam, Majestic, Oct 26, 2026
- Earned media still drives more than 84% of AI citations, Muck Rack via GlobeNewswire, Oct 28, 2026
