Your Rank Tracker Broke for Three Weeks
SERP collection broke for 20+ rank trackers during the September spam update window. Some ranking drops were data failures. Here is how to tell.
Some of the rankings you "lost" in September were never lost. Your tracker just stopped looking.
Between roughly September 17 and October 1, Google changed how search results get delivered to automated systems. The change arrived alongside the September 2026 spam update, and it quietly broke SERP collection across the rank tracking industry. Trackers that were supposed to collect a full page of results sometimes came back with one to three listings. A keyword sitting at position 38 got reported as missing because the crawler only reached position 20. Nothing moved in Google. The measurement did.
The scale only became public this week. Ranktracker published its technical investigation on October 7, then announced on October 8 that it had tested more than 20 competing platforms and found incomplete SERP collection across the tested platforms. Its own infrastructure was restored on October 1. Other providers, the announcement says, are still catching up.
If you report rankings to a client or a boss, this should worry you. Not because the tooling had a bug, but because the bug looked exactly like the news.
The outage hid inside the update
The September 2026 spam update was the longest spam update in recent memory. Google confirmed the rollout completed on October 8, with three visible waves (September 25 to 27, September 30, and October 4 to 6). The three previous spam updates this year each finished in under three days. This one took the full two weeks Google had warned about, per the timeline eology documented.
That long window is what made the tracking failure dangerous. Ranktracker first noticed unusual SERP behavior around September 17, a full week before Google announced the spam update. As the update rolled out, the collection problems got worse. So for roughly three weeks, real Google volatility and fake tracker-created volatility were layered on top of each other, and the dashboard had no way to tell you which one you were looking at.
Glenn Gabe said it plainly in his update coverage: Ahrefs, Semrush and other rank trackers were showing unreliable SERP data, and Google Search Console mattered more than third-party tracking during this period. During that window, a rank tracker dashboard was not a measurement. It was a suggestion.
The failure mode matters too. When a tracker collects only part of a SERP, the missing positions do not show up as blank cells with a little "collection failed" note. They show up as losses. The keyword is marked lost, outside the top 100, or down sixty positions. A campaign that was perfectly healthy could look like it had suffered a major decline. Multiply that by thousands of keywords and you get distorted visibility scores, wrecked position distribution charts, and client reports that need retractions.

How I check whether tracker data is poisoned
I run SEO reporting at First Rank, so I live inside these dashboards. When tracking data looks suspicious, I run the same protocol every time. It is built around one rule: never trust a single number during a known tool outage. Here is the protocol.
1. Quarantine the window. Any ranking movement reported between September 17 and October 1 gets flagged as unverified until proven otherwise. I annotate that window in the report itself. Clients remember what you told them in October when they read November's numbers.
2. Hunt the disappear-reappear pattern. Real ranking losses look like a decline curve. Collection failures look like a cliff: a group of keywords vanishes on the same day and comes back together. Ranktracker's investigation describes exactly this signature, a keyword at position 32 that ranks, disappears, then reappears across four days while the page never moved. If I see that shape, I stop diagnosing the site and start diagnosing the data.
3. Anchor on Search Console. Impressions and clicks from GSC are Google's own measurement. They kept working through the whole period. In my ecommerce SEO work, the campaign I care about most is the one that grew organic clicks 311 percent. That growth number comes from clicks, not ranks. When the tracker said one thing and the clicks said another, the clicks were right every time. I wrote a longer argument for this in Rankings Are Not Revenue: report the metric the business gets paid on first, and treat ranks as a supporting metric, which is how I report it on my own site too.
That last point is the one I keep coming back to. The agencies that panicked in September were the ones whose entire reporting stack was a rank tracker. The ones that stayed calm had click and impression data underneath. The outage tested what your reporting is built on.
The question your tracker should answer
Ranktracker's investigation ends with the right question, so I am going to steal it: the question should not be "does my rank tracker show my rankings." It should be "how does my rank tracker behave when Google does not return the expected SERP."
For every tool you pay for, this matters. A tracker that retries incomplete searches is telling you about its failure handling. A tracker that does not is giving you a clean dashboard built on top of an unreliable crawl. One of them will cost you a confused week. The other will cost you a client conversation.
If you ran a site review or a client report during the September window, do not quietly move on now that the data looks normal again. Go back to the lost-keywords list from that period, especially anything that ranked between positions 20 and 100, and manually verify whether those keywords are gone. Some of them will turn out to be alive. I covered the update itself in my September spam update explainer, and the real losses from that update are the ones worth your attention, not the phantom ones.
One more thing for the agencies. If you sent a report during the outage that showed ranking drops, send a correction. "Our tracking tool was returning incomplete data between these dates" is an awkward email and a necessary one. The alternative is a client who made decisions based on numbers that were never real.
A rank tracker that cannot say "I failed to collect the SERP" will say "you lost the ranking" instead. One of those is a bug you can fix. The other is a report your client believes.


About the author
Dalin de Graff is an SEO and GEO Strategist at First Rank, where he built the agency's GEO practice. He writes about search, AI, and marketing at dalindegraff.com.