Your Rank Tracker Just Lost Its Eyes. What to Trust Instead.

Rank tracking accuracy has gotten worse, and it isn’t your SEO team’s fault. Google has changed how outside tools can see its results three times since January 2025, and each change makes third-party position data a little less trustworthy. If a dashboard says your keywords fell but your organic traffic and leads didn’t, the measurement may have moved while your performance stayed put.

That matters because plenty of agencies and in-house teams still report rankings as if they were hard facts. They aren’t anymore, at least not beyond the first page. Below we cover what changed, what the best public case study shows, which numbers to be skeptical of, and what to measure instead. This is core measurement work in our AI audits, because you can’t judge visibility with an instrument you can’t trust.

Why did my keyword rankings drop in my tracker when nothing changed?

Often because Google changed what the tracker is able to see. Rank trackers don’t read Google’s internal data. They load search pages the way a person would and record what’s there, so every time Google makes that harder, the report changes even if your pages didn’t.

Dan Lauer of Walker Sands laid out the sequence in Search Engine Land on September 24, 2026. In January 2025 Google began requiring JavaScript for search, which caused data blackouts at several tools, including Similarweb, Rank Ranger, SE Ranking and ZipTie. Ahrefs and SISTRIX were reported as unaffected. In September 2025 Google stopped supporting 100 results per page. And in August 2026 it rolled out google.com/goto passthrough links on search results.

What happened when Google removed num=100?

Most tracked sites suddenly looked smaller. Tyler Gargula of LOCOMOTIVE Agency analyzed 319 properties and published the results on September 18, 2025. He found 77.6% lost unique ranking terms and 87.7% saw fewer impressions in Search Console. Queries also shifted toward the top three and page one, with fewer showing up on page three and beyond.

Gargula’s own framing is the useful part: fewer impressions and keywords can make reports look weaker even when rankings hold steady. A big share of those deep-page impressions came from bots and scrapers pulling 100 results at a time, not from people. Semrush and AccuRanker both acknowledged disruption at the time, and PPC Land reports AccuRanker limited tracking to the top 20 results afterward.

What is Google’s /goto change, and why does it hurt rank trackers?

It’s a redirect layer. Instead of linking straight to your page, a result link now points to a google.com/goto address that sends the visitor onward, and a scraper can’t read the destination from the page. Barry Schwartz at Search Engine Roundtable reported on August 26, 2026 that Google confirmed the rollout. A Google spokesperson said the company regularly takes steps against evolving forms of abuse, and the coverage framed the target as scraping by outside services, including SerpApi, which Google sued in December 2025.

The cost falls on everyone downstream. PPC Land, citing Nozzle’s Derek Perkins, reports that resolving all the links for a single five-page set of results can take 500 to 1,000 requests, where reading the page used to need none. Perkins’s measurements put the rollout at close to 100% across several residential IP providers. Nothing in the coverage we read says Google commented on what this means for rank tracking or Search Console, so treat the downstream effect as the industry’s reading, not Google’s statement.

Which rank tracking numbers should you stop trusting?

Start with anything past position 10 to 20. In Lauer’s Reddit example, the tracked keyword count dropped about 25% (roughly 60 million keywords) and tracked SERP features dropped about 21% (roughly 13 million) in May and June 2026, while Reddit’s organic traffic stayed flat. The deep positions fell hardest: down 28% at positions 11 to 20, 34% at 21 to 50, and 35% beyond 50. Position one went up 16% and the top three went up 11%.

Read that as a pattern: the lower you go, the less the tool can see. Here’s what we’d treat with caution from now on.

• Total keyword counts and “visibility” scores. They depend on deep-page data that tools can no longer collect cleanly.
• Anything dated across a break. A year-over-year chart that crosses September 2025 or August 2026 is comparing two different instruments.
• Share-of-voice dashboards. They’re built on the same scraped positions.
• Competitor keyword totals. You can’t verify them, and neither can the vendor, fully.

Top-of-page positions for a short list of keywords you care about are still reasonably useful. Just don’t hand a client a 5,000-keyword trend line and call it truth.

How do you measure SEO without relying on rank tracking?

Use data you own, and let rankings play a supporting role. Lauer’s suggestion is to look at how often organic is the first touch, how organic visitors convert, and whether other channels move when organic visibility moves. That’s our experience too, and here’s the stack we’d put in place.

1. Search Console for clicks, not impressions. Impressions took the hit from num=100. Clicks and click-through rate on your priority queries are closer to real behavior.
2. GA4 or your analytics tool for organic sessions, conversions and first-touch. Annotate September 2025 and August 2026 on every chart so nobody reads a step change as a performance problem.
3. CRM data for leads and revenue by source. This is the number your CEO actually cares about.
4. A short, fixed list of tracked keywords. Keep 30 to 50 that map to revenue and check only their top-of-page positions.
5. Repeated AI answer sampling. Ask the same buying questions in ChatGPT, Perplexity and Google’s AI Mode on a schedule. We cover how to build that question list in our guide to building an AI prompt set, and how to tie it to revenue in our piece on measuring AI search ROI.

What should you tell your client or boss when the tracker drops?

Tell them what was checked, in this order. First, did clicks, conversions and revenue from organic change? If not, the drop is probably in the tool. Second, did the drop line up with one of the three Google changes above? Third, did the loss sit in deep positions rather than the top ten? If the answers are no change, yes, and yes, you have a measurement story, not a ranking story.

Document it once and reuse it. A one-page note that lists the dates, the first-party numbers, and the tracked positions that actually matter will save you from explaining the same red arrow every month. If the first-party numbers did fall, that’s a different conversation, and a real one.