Why your rank tracker shows strange drops, and how to tell a real one

Often it is bad data. Rank-tracking tools have been receiving normal-looking Google results pages with the wrong sites in them, so before you act on a drop, check that it lasted beyond a few days, that it isn't only on page 2, and that Search Console and your enquiries show the same fall.

If your rank tracker has lately shown your site falling out of Google's top 10, the tool may have been handed a bad page, and you may not have lost the ranking at all. Nozzle, a rank-tracking company, reports that since 13 September, 27.6% of its checks have come back poisoned on page 1 or 2, against 0.28% before. On those poisoned pages, a tracked project's own site misses the top 10 on 63.7%, against 2.6% normally.

Put those two figures together and roughly one tracker check in six could have shown a tracked site out of the top 10 for no real reason: 27.6% multiplied by 63.7% comes to about 17.6%. That is an average across Nozzle's data, not a reading of your site, and the daily rate swings a long way. The worst day so far was 2 November, at 68.6%.

So a drop in a rank tracker is not, on its own, proof of a lost ranking. Tools are being served pages that look normal but have the wrong sites, most last for one fetch, and the tool stores them as real positions.

At JezNorthWeb, we think a daily position number was always a noisy guide, and right now it is noisier than usual. A second thing is moving results this month as well: Google's November core update. Here is how to tell bad data from a real change, and what to check before you touch your website.

How often are rank trackers getting bad pages?

The trouble started in mid-September. According to Nozzle, requests to Google started failing on 13 September, and from 20 September the ones that got through came back "poisoned". Before then the problem barely existed. It now runs at about 100 times its old rate (27.6% against 0.28%).

Tracking a phrase locally doesn't protect you. Search Engine Journal's summary of Nozzle's data found local targeting made no difference (7.5% against 7.4% for the same phrase tracked nationally and locally). A tracker set to Evans Head or Lismore is in the same boat as one set to the whole country.

Australian searches weren't spared either. TrueRanker, another tracker vendor, passed on Nozzle's figures up to 9 October and listed Australia at 11.5% poisoned, ahead of the US at 8.5% and the UK at 3.0%. Those figures are older and second-hand, so treat them as a sign that Australia was hit, not as today's rate.

Who is doing this, and why, is not settled. Nozzle's Derek Perkins describes a three-part Google approach: very wide IP blocking of residential proxy networks, many more captchas, and bad data injected into suspect results pages. Search Engine Roundtable's Barry Schwartz reads the target as AI bots and a results-scraping service rather than SEO tools themselves. Google hasn't commented publicly. Everything here is the vendors' and reporters' reading, not anything Google has confirmed.

What does a poisoned results page look like?

A blocked check is easy to live with. The tool gets nothing back, and you see a gap in the report.

A poisoned check is the problem. Nozzle describes complete, normal-looking pages whose results have been swapped out: "Nothing on the page gives it away." The tool can't tell the difference, so it records whatever position it finds.

Page 2 is where it mostly happens. Nozzle found page 2 is poisoned 1.5 times as often as page 1, "mostly under a page 1 that looks fine". Social sites, Facebook most of all, take the place of the usual sites, and a rank report reads it as a crash.

Semrush describes two forms. In one, the whole page is off, even on page 1. In the other, the page looks normal at first but includes pages lower down that only loosely match the search.

The scale of the swap is what makes it look like a disaster on a report. Search Engine Journal's summary says poisoned results keep only 20.7% of the websites from the previous search, against 88.5% normally. Most of the page changes at once, so your site seems to vanish along with nearly everyone else's.

How long does a fake drop last?

Usually not long. Of poisoned spells, 80% last one search result, 15% last two and 5% last three or more. A drop that shows on one check and is gone on the next is the commonest pattern.

Recovery on a given phrase can take longer, though. Nozzle measures how soon the usual sites return: only 45.1% are back within three days (67.9% through 21 October), and 89.0% within two weeks. So waiting three days is a good first test, but a drop still showing after three days isn't automatically real. That is why Search Console is the deciding check, further down.

Which dates in your reports deserve a second look?

If you keep monthly reports, mark these dates. Anything in your tracker from 20 September on could include poisoned results. One tool told Search Engine Journal it re-processed keywords affected by incorrect results (25 to 28 September and 9 to 13 October), which gives you a sense of the worst-affected spells for that tool's users.

The problem hasn't neatly ended. As of 13 November, TrueRanker says it is ongoing, and poisoned results still occasionally appear from page 2 onwards. Other vendors report different levels, so no one can say trackers are fixed, or broken, across the board.

Is this drop real? Four questions before you act

Work through these in order. Stop as soon as one answer says "bad data".

  1. Did it come back within a few days? If it did, it was very likely a poisoned check.
  2. Did only page 2 move? A drop that sits only on page 2 matches the pattern Nozzle describes.
  3. Does Search Console agree? Look at the same query, the same dates and the same device. Semrush's advice is: “If you see unexpected ranking changes, especially for keywords unrelated to your pages, compare against your own data, such as Google Search Console.” If Search Console still shows clicks and impressions, the site didn't disappear from Google.
  4. Did you lose enquiries or calls? This is the one that matters to your business. If the phone and the inbox are steady, a red line on a report is not an emergency.

TrueRanker's own rule of thumb puts it in one line: treat a drop as real only if it stays, Search Console agrees, it didn't come back within three days, and it isn't only on page 2. That is one vendor's rule, not Google's, but it is a sensible order to check things in.

Is the November core update moving rankings as well?

Yes, and that is what makes this month confusing. Google called the update that began on 10 November “a regular update designed to better surface relevant, satisfying content for searchers from all types of sites”, and Search Engine Roundtable says it “is not a penalty”. Google said the rollout “may take up to 2 weeks to complete”. When last checked, Google's Search Status Dashboard listed it from 10 November 2026 with no end date logged.

Core updates aren't rare. The same Search Engine Roundtable report notes Google typically runs 2 to 4 core updates a year, with a median rollout of 14 days.

Until this rollout ends, real movement and tracker noise overlap. A drop that holds past a few days, shows on page 1 and appears in Search Console for the same query deserves attention. Even then, our advice is to note it and wait for the rollout to finish before rewriting pages. Rankings often keep shifting until a core update settles, and changes made mid-rollout are hard to judge.

Can you trust Search Console when the tracker looks wrong?

Mostly, but it has had its own oddities lately, so know which dates to skip.

The practical rule is to compare like with like: the same query, the same date range and the same device, and avoid weeks with a known glitch in them.

What should you ask your rank tracker or SEO agency?

Tools have handled this differently, and they say so. Search Engine Journal noted that as of 22 October none of the vendor statements included a statement from Google, and none cited independent testing of how well their checks work. Their fixes are their own judgement, so it's fair to ask how they work.

Old reports are the easy thing to miss. Search Engine Journal suggests you ask the vendor whether stored data for those dates was updated, because older exports may not match the tool now. A PDF you downloaded in October may show a crash that the tool itself has since cleaned up.

If someone sends you a report showing a sudden collapse, the most useful question is the simplest one: is this built on one day's data?

What tells you search is actually working?

For a small business, the measures that count are the ones that reach your bank account: enquiries, calls, bookings and the clicks Search Console records. A position number is a guide to those, not a substitute for them.

Pick a few phrases that bring in work, check them in Search Console once a month over the same date range, and set that next to how many enquiries came in. That gives you a steadier picture than any daily chart. Our guide to measuring WordPress website performance covers the other numbers worth keeping an eye on, and our search engine optimisation page explains how we approach search for local businesses.

Before you change your site over a report, send it to us

A drop that came back, sat on page 2, or doesn't show in Search Console usually needs nothing from you. A drop that holds, shows in Search Console and lines up with fewer calls is worth a proper look, especially while the core update settles.

If a rank report has you worried, send it to JezNorthWeb through our contact page before you rewrite anything. These three things help us read it quickly:

Sources

Recent posts

Ready for a new website?

Let’s create a web design that promotes your business and attracts new clients

We help you every step of the way with design, domains, hosting, email accounts and website support.

Contact us now to arrange a free consultation about your website

Make an AppointmenT