How accurate are rank trackers after Google's swap?
Quick answer: Rank trackers are still mostly accurate, but you shouldn't take every drop at face value. Since mid-September 2026, Google has been sending some scrapers results pages with the listings swapped out, so a tracker can show a site falling out of the top 10 when nothing has changed. The damage is limited. It's worst in deeper positions, and most errors clear up within days. Before you act on a drop, check it against Search Console impressions, clicks and average position.

Put your organic marketing on autopilot
artiql researches, writes and publishes SEO + GEO content in every language — and turns each article into a video. See it run on your brand.
What did Google actually do to rank trackers?
Since mid-September 2026, Google has been sending some automated scrapers, rank trackers included, results pages that look normal but have most of their organic listings swapped out. There's no error message and no CAPTCHA, so the tools save these pages as real data, and a site can seem to drop out of the top 10 overnight. SEOs now call these "poisoned SERPs." One rank-tracking vendor checked 17.8 million results pages across 49 countries. Between September 13 and October 9, 8.0% came back poisoned, up from 0.17% before. The worst day was September 29, at 27.0%, and since early October the daily rate has stayed somewhere between 5% and 11%.
It happened in stages. Starting September 13, requests to Google began to fail outright. From September 20, many of the requests that still got through came back with swapped results. Google has also been blocking large ranges of IP addresses tied to residential proxy networks and showing a lot more CAPTCHAs, and several tracking tools have reported missing positions in recent weeks. Google hasn't commented, and it hasn't confirmed a motive. Its spam policies already ban automated queries for rank checking without permission, though. The most common theory is that Google is going after bulk scraping that feeds AI answer engines, and rank trackers collect results the same way, so they're getting caught too.
Why should a business owner care? Because this is how most of you judge whether your SEO is working. You publish content for months, open your tracker on a Monday, and see your best service page has dropped from position four to nowhere. In that same dataset, poisoned pages kept only 20.7% of the sites that ranked the day before, compared with 88.5% normally. A tracked site missed the top 10 on 58.6% of poisoned pages, versus 2.7% on clean ones. Charts like that lead to panicked emails to your agency, rewrites of pages that were doing fine, and sometimes cancelled contracts. You don't need to throw out your tracker. You need to know which of its numbers to trust.
How much does this skew your Semrush or AccuRanker data?
For most businesses, much less than the headlines suggest. Semrush has said publicly that Google has been sending bad results to automated collectors since late summer. It also says the effect on tracked rankings has been limited, because the problem has mostly shown up in deeper results rather than the page-one positions most people watch. Semrush has added validation checks and keeps adjusting them as Google changes tactics. So if you suspect your Semrush ranking data is wrong, look at keywords sitting on page two or three first. A term bouncing between positions 18 and 45 is far more likely to be bad data than a keyword that has held third place for months.
AccuRanker's figures point the same way. It saw a sharp spike in late September that has since fallen to about 1% of its keyword checks. Google tightened detection again on October 15, mostly on mobile, and both mobile and desktop were back under 1% by October 17. AccuRanker also found that Google seems to decide per browser session, not per keyword, so one bad session can corrupt a batch of checks at once. And when it went back through its stored rankings, it found smaller waves of the same pattern from late January, with local searches hit first. If you track map-pack terms like "dentist near me," those deserve the most caution.
Much depends on how a tool reaches Google. During one ten-day stretch, one collection route returned 1% to 7% poisoned pages while another returned 9% to 30% on the same days. That's why two trackers can disagree about the same keyword on the same morning. Most errors also don't last: in the largest study, 77.9% of sites that went missing on a poisoned page were back within three days. My view is that a one-day drop in any tracker has never been worth acting on, and now it's even less meaningful. Look at the three-day or weekly trend before you change anything on the site.
| Source | What was measured | Reported rate |
|---|---|---|
| Large-scale vendor study | Share of judged SERPs, September 13 to October 9 | 8.0% (peak 27.0%) |
| Same study, since October 1 | Daily share of poisoned SERPs | Roughly 5–11% |
| Second tracking provider | Daily SERPs affected, undisclosed method | About 11% |
| AccuRanker | Keyword checks after the late-September spike | Around 1% |
Why do my rankings look different in Search Console than in a rank tracker?
Even before any of this, Search Console and rank trackers didn't measure the same thing. A tracker takes one snapshot of one results page, usually once a day, for a single location and device, with no personalization. Search Console doesn't scrape anything. It reports what happened in real searches: how often your pages appeared, how often people clicked, and where you ranked when they saw you. So the question "why do my rankings look different?" usually has a simple answer. The two tools are answering different questions, so they'll rarely give you exactly the same number, even when both are working perfectly.
Average position is where most of the confusion starts. For each impression, Search Console records the highest position any of your pages held, then averages that across every impression in your date range. If a clinic's page ranks third for people searching nearby and twelfth for everyone else, its average might land around seven or eight. A tracker set to the city center would show three. Neither number is wrong. Low-volume queries also swing a lot, because a handful of impressions from an unusual location can move the average sharply. And to protect privacy, Search Console leaves out very rare queries, so long-tail totals won't always add up.
Here's the part that matters right now: poisoned SERPs only affect the scraped snapshot. Search Console comes from Google's own record of real searches by real people, so swapped results served to bots never appear in it. That makes it the natural referee when your tracker and reality disagree. It has its own limits. Data usually runs a day or two behind, and average position can hide big differences between devices or locations unless you filter for them. Still, if your rank tracker accuracy is in doubt, Search Console is the most reliable data you have about your own site, and it costs nothing.
| Attribute | Rank tracker | Search Console |
|---|---|---|
| Data source | Scraped results pages | Google's own logs of real searches |
| Location and device | One fixed setting per check | Every location and device combined, filterable |
| Position metric | Exact position in one snapshot | Average of your highest position per impression |
| Affected by poisoned SERPs | Yes | No |
| Shows clicks and impressions | No, only estimates | Yes, from actual searches |
| Competitor visibility | Yes | No, your own site only |

Put your organic marketing on autopilot
artiql researches, writes and publishes SEO + GEO content in every language — and turns each article into a video. See it run on your brand.
How do you check whether a ranking drop is real?
Treat every drop your tracker shows as unconfirmed until Search Console backs it up. Start by writing down three things: the keyword, the page that used to rank, and the date of the drop. Open the Performance report in Search Console, filter by that query (or by the page), and compare the days around the drop with the period before. Look at impressions, clicks and average position together. If impressions and clicks held steady while your tracker said you'd disappeared, nothing happened to your site. You almost certainly got a swapped results page. Note it and move on. A real drop shows up as falling impressions and a worse average position across several days.
Next, look at how wide the problem is. A real algorithm hit or technical failure almost never touches just one keyword. It shows up across a group of related queries, often for a whole section of the site, and it shows up in Search Console too. If one keyword fell out of the top 10 while twenty similar terms stayed put, that points to bad data. Give it three days, because most poisoned results clear up on their own. Then check your analytics for organic sessions and leads on the affected pages. Revenue doesn't lie, and a page that's still bringing in enquiries isn't the emergency your dashboard makes it look like.
If you're an agency or manage several sites, automate this step. The Search Console API lets you pull daily clicks, impressions and position for your key queries and line them up against tracker data. Then any tracker drop without a matching dip in Search Console gets flagged automatically before a client ever sees it. Search Console keeps 16 months of performance history, which is plenty to compare against last year's seasonality. One more tip: hold off on page changes for at least a week after any unexplained drop. Rewriting a page that was never really hit is the most expensive mistake here, because you might break something that was working.
What should SEO reporting look like in 2026?
Good SEO reporting in 2026 leads with your own data, meaning clicks, impressions, conversions and AI citations, and uses tracked rank positions as supporting context. I'll go further: poisoned SERPs are a useful wake-up call. Daily positions for a fixed keyword list were always a rough proxy, made worse by personalization, local results, AI Overviews and endless SERP features. Owners and agencies came to treat a position number as the result itself. It never was. The result is qualified visitors and enquiries. If your monthly report opens with a table of positions instead of organic clicks and leads, it's measuring the wrong thing, poisoned results or not.
Rank trackers still earn their place. They're the only practical way to watch competitors, spot new SERP features and track keywords before they get enough impressions to show up reliably in Search Console. Use them for trends: look at weekly averages, group keywords by topic, and look at the share of terms in the top 10 rather than individual positions. Pair that with Search Console for the truth about your own site, then translate the gains into business terms. To see what moving up a few spots could mean in visits, plug your keywords into our SEO traffic calculator and compare the estimate with the clicks you're actually getting.
There's one blind spot that Search Console and rank trackers both share: AI answers. When ChatGPT, Perplexity, Gemini or Claude cites your page, neither tool reliably shows it, yet those citations are fast becoming a second organic channel. A quick GEO score check tells you how ready your pages are to be quoted. If you're weighing organic content against your ad budget, our breakdown of whether SEO is cheaper than Google Ads in 2026 walks through the cost-per-lead maths. Artiql publishes search-ready content on autopilot and reports on the numbers that matter, with pricing tailored after a short discovery call. Book a demo to see how it would work for your site.
Frequently asked questions
Is Google deliberately feeding Semrush and other trackers wrong data?
Google hasn't confirmed anything or commented on the research. What's documented is that since mid-September 2026, some automated requests have received complete-looking results pages with swapped listings, along with more IP blocking and CAPTCHAs. The leading theory is that Google is fighting bulk scraping for AI tools, and rank trackers get caught in it. Google's spam policies already prohibit automated rank-checking queries without permission.
Do poisoned SERPs affect my real Google rankings?
No. The swapped results only go to the bots scraping Google, not to real people searching. Your actual visibility, traffic and leads don't change at all. What changes is the picture your tracker shows you. That's why Search Console is the right place to check: it's built from real searches, so if impressions and clicks are steady, your rankings are fine.
Should I cancel my rank tracker subscription?
For most businesses, no. Trackers are still the best way to watch competitors, spot SERP features and follow keywords too new to show up in Search Console. Major vendors report the problem is limited, often around 1% of checks and mostly in deeper positions. Just change how you read the data: follow weekly trends, ignore one-day drops, and confirm anything worrying in Search Console.
How quickly does Search Console show a real ranking change?
Search Console performance data usually runs a day or two behind real searches, so you can confirm or rule out a suspected drop within about 48 hours. To be confident, compare at least several days on both sides of the drop, since one day of low-volume data can swing a lot. If impressions and average position hold steady across that window, the tracker drop was almost certainly bad data.

Put your organic marketing on autopilot
artiql researches, writes and publishes SEO + GEO content in every language — and turns each article into a video. See it run on your brand.