Sean Walsh
Sean is a Director at Intelligency heading up our digital marketing and client services operations. Sean has 15+ years experiencing working both in-house and agency with brands including Lloyds, Alstom, Hitachi, Lufthansa, Viaplay, DFDS Seaways and Mercedes-Benz.
If you have looked at a rank tracking report this autumn and felt your stomach drop, you are not alone. Keywords that sat comfortably on page one for months have vanished overnight, then reappeared a few days later as if nothing happened.
In a lot of cases the website hasn’t changed. What has changed is how reliably anyone can measure Google from the outside. Rank tracking has always been an estimate, but 2026 is the year that estimate became noticeably less dependable, and marketers need to adjust how they read the numbers.
Google is making it harder for tools to see what you see
Rank trackers work by sending automated searches to Google and recording where your site appears. Google has never liked this. Its own spam policies list “scraping results for rank-checking purposes” as machine-generated traffic, and over the last two years it has steadily closed the doors.
In January 2025 Google started requiring JavaScript to load search results, which made basic scraping more expensive. In September 2025 it stopped supporting the “&num=100” setting that let tools pull 100 results in one request, which is why so many Search Console accounts showed impressions falling and average position “improving” overnight. Bots had been generating a chunk of those impressions all along. Then in the summer of 2026 Google began wrapping result links in redirect URLs, first tested in July and rolled out in August, adding another layer tools have to unpick.
None of that was aimed at SEO agencies specifically. The bigger target is the wave of AI companies and data resellers scraping Google at scale. But rank trackers use the same methods, so they are caught in the crossfire.
The bigger problem: results that look real but aren’t
The most worrying development came in September. Rank tracking provider Nozzle published research showing that, from 13 September, Google appeared to be serving deliberately altered results to traffic it suspected was automated. Rather than blocking the request, it returned a page that looked normal but wasn’t.
Nozzle has kept updating the figures. As of 1 November, 26.9% of page one and two results it collected had been affected since 13 September, against 0.28% beforehand. On page one alone the figure was 14.3%, and it peaked at 64% on 31 October. On affected pages, a tracked site fell out of the top 10 63.5% of the time, compared with 2.6% normally. Just under half returned to normal within three days.
That explains the pattern many of us have seen: a sudden, dramatic drop across lots of keywords at once, often followed by an equally sudden recovery, with nothing in analytics to back it up.
Google hasn’t confirmed any of this, and the cause isn’t certain. Other explanations, such as heavier localisation of results and ranking volatility after the May 2026 core update, are also in play. But the tool providers themselves aren’t hiding from it.
What the rank tracking providers are saying
Several of the main providers have published status updates or support notices. The fact that they are saying it publicly is useful in itself. It confirms this is an industry-wide problem rather than a fault with one tool or one client’s site.
- SISTRIX (support notice, updated 7 October): has seen divergent results since September. It now only publishes rankings that pass multiple checks and re-collects any that fail, so some keywords update less often. On-demand updates have been switched off. It maintains that top 10 rankings and its Visibility Index remain reliable.
- SE Ranking (X post on 16 October, plus its status page): has made several fixes and corrected ranking data for the affected dates, but said the issue still affected roughly 2% of tracked results.
- AccuRanker (blog, from 19 October): on its worst day, 24 October, 40% of its first desktop checks came back altered. After re-checking suspect pages with a fresh session and comparing each result against the keyword’s recent history, 3.2% of stored results still failed. By 25 October that was down to around 1%.
- Ahrefs (blog, 22 October): is seeing between 5% and 20% of results affected depending on the data provider. It has added filters to reject abnormal results and is cleaning up affected historical data.
- DataForSEO (blog, 23 October): this data provider supplies many other tools. Its CTO estimated 20% to 30% of scraped results were affected from 26 September. Even with two retries and historical checks, 2% to 3% still slip through. It advises annotating reports from 20 September and treating one-day drops as unconfirmed.
- Semrush (blog, 27 October): describes it as an industry-wide issue. It has added validation checks that flag and re-collect suspect results, and asks users to contact support with the domain and keywords if something looks wrong.
This is an ongoing cat-and-mouse game, not a one-off glitch. Fixes go in, Google changes something, and the numbers move again. Some data providers saw failure rates jump again on 26 October, and Nozzle’s peak came on 31 October.
Search Console isn’t perfect either
The natural response is to lean on Google Search Console instead, and for the most part that’s the right call. It is Google’s own data on real searches by real people. But it has its quirks. Between 31 October and 3 November, some sites saw an unexplained spike in clicks in the Performance report, which some in the industry suspect was bot activity. Google hadn’t commented at the time of writing.
Average position in Search Console also needs care. It is an average across every query, device, location and result type over the period selected, not a snapshot of where you sit for one keyword on one day. It will rarely match a rank tracker, and it isn’t meant to.
What this means for your reporting
Rank tracking still has value. It is the quickest way to see broad direction across a keyword set and to keep an eye on competitors. But a single-day ranking drop is no longer evidence that something is wrong with your site. Acting on it can do more harm than good, especially if it leads to rushed content changes or unnecessary panic with senior stakeholders.
This is how we are adjusting our approach:
- Trust the trend, not the data point. Look at visibility and ranking ranges over weeks, not daily positions.
- Check Search Console before reacting. If clicks and impressions are holding while the tracker shows a collapse, the tracker is the likely problem.
- Watch for whole-page swaps. If several usually stable rivals vanish from the same results page at once, treat it as suspect data.
- Wait three days. Altered results typically correct themselves quickly. No content changes on the back of a tracker drop alone.
- Check your provider’s status page. Before raising a drop with a client, see whether your tool has published a notice for the same dates.
- Keep a few reference keywords. Track a handful of stable, well-understood searches. When they behave strangely, everything else probably is too.
- Track locally where it matters. For clinics and multi-location businesses, city-level tracking is far more meaningful than a national average.
- Put business outcomes first. Enquiries, bookings, sales and leads are the numbers that matter. Rankings are a supporting signal, not the headline.
Google has become much stricter with the automated tools that check rankings, and some of those checks are now returning unreliable results. Our tracker is still useful for spotting trends, but we’re validating any big movement against Google’s own Search Console data and, most importantly, against enquiries and sales before we act on it. The wider lesson is that search measurement is getting messier on every front, from AI Overviews to ChatGPT referrals to rank tracking itself. The businesses that cope best will be the ones that judge SEO on the outcomes it drives, not on a single number in a weekly report
