Rank tracking: what a #1 position no longer promises

The number that makes "rankings are rankings" wrong

Rand Fishkin published the analysis through SparkToro on 9 June 2026, built on Similarweb's desktop and mobile web clickstream panel for US searches from January through April 2026. Zero-click searches, meaning the user never left the results page, reached 68.01%, up from 60.45% across the same measure in 2024. Clicks to any destination, organic results, ads, or Google's own properties, fell 9.51 percentage points over that period, a 22.9% decline in searches that generate a click at all. AI Overviews showed on more than one in five queries in the panel, and on the queries where one appeared, click-through fell by nearly 60% against queries without one.

That last figure is the one that changes what tracking needs to do. A page holding position one under an AI Overview and a page holding position one without one are not delivering the same result. A rank tracker that only records position shows them as identical anyway. Flagging AI Overview presence per query, which is what a subscription like managed rank tracking is built to do, turns that invisible difference into something you can watch happening to a specific keyword.

Why the average position in your dashboard can look fine while the traffic isn't

Dan Taylor made the mechanical case for why in Search Engine Journal on 17 July 2025. Search Console, and most tools that inherit its logic, blend every placement type into a single average. If a page shows as an AI Overview's top citation, effectively position one, on the same query where its plain blue link sits at position four, Search Console reports an average position of roughly 2.5, which reads as a strong page-one presence. In reality, most of the traffic that query is ever going to send comes from whichever of those two placements a searcher actually clicks, and the zero-click data above says a meaningful share of the time neither one gets clicked at all. Taylor's recommendation is to stop treating average position as the headline number and instead track click-through by feature type, actual traffic volume, and conversions, the instruments that show what a searcher did rather than what the algorithm decided to blend. Our guide to measuring AI visibility covers the fuller instrument set, including the ones Search Console cannot give you at all.

This is the specific failure mode a plain position number cannot catch on its own. A tracker that reports "position 2.5, up from position 4" looks like good news. Whether it actually is depends entirely on which placement is doing the work, and that requires seeing the AI Overview flag sitting next to the position, not folded into it.

And the positions themselves are moving faster than most tracking schedules

SE Ranking's Yulia Deda published a comparison of Google's March 2026 core update against December 2025's on 17 April 2026, built from 100,000 keywords across 20 niches, tracked before and after each update from New York, USA. The March period combined a Spam Update running 24 to 25 March with a Core Update running 27 March through 8 April. Movement in the top 3 results rose from 66.8% of keywords in December to 79.5% in March. Top 10 movement rose from 83.1% to 90.7%, and top 100 movement from 96.9% to 98.5%, essentially the entire tracked set. More specifically, 24.1% of pages holding a top 10 spot before the March updates dropped out of the top 100 entirely afterward, up from 14.7% in December, and nearly 30% of the pages that ended up in the new top 3 had been ranked outside the top 20 beforehand, more than double December's 13%.

A quarterly check, or a check that only happens after someone notices traffic fell, cannot see any of that. A page can enter the top 3, sit there for two weeks, and drop back to page two before the next scheduled look, and the record shows no story at all, just two data points that happen to look similar. That is the same blind spot our guide to diagnosing a traffic drop keeps running into: without a continuous line, ranking volatility and a real ranking loss are impossible to tell apart after the fact.

What tracking should measure this year

Three changes follow directly from the data above. First, record AI Overview presence as its own field next to the position, not blended into it, since the same reported number can mean two different outcomes depending on which placement it came from. Second, check weekly rather than monthly. With top 3 movement running at nearly 80% between two consecutive core updates, a four-week gap between checks is long enough to miss a keyword's entire cycle in and out of a placement. Third, track both Google and Bing, not just the engine assumed to matter most. Bing's index feeds Microsoft Copilot's answers directly, so a position shift there can predate a citation change before it ever shows up in a Google-only report, the same reasoning our guide to winning an AI Overview citation applies to Google's own answer layer.

None of this argues against the concept of rank tracking. It argues against treating a single blended monthly number as if it still tells the whole story it told five years ago, before AI Overviews existed and before the SERP moved this much between algorithm updates.

What we could not verify

Fishkin's own methodology note excludes Google's mobile app, where he writes that zero-click features operate "even more aggressively" than on the mobile web, meaning the 68.01% figure likely understates the true zero-click rate for anyone searching primarily inside the app rather than a browser. Taylor's 2.5 blended-position example is illustrative arithmetic showing how the calculation works, not a figure pulled from a live Search Console export, and the exact blend in any real case depends on the actual impression volume at each placement. SE Ranking's volatility figures come from one geography, New York, USA, and 20 tracked niches; a specific site outside those niches or that market is not guaranteed to see matching swings, even though the direction, more volatility in March than December, is consistent with what other trackers reported the same month.

A check worth running this week

1. Pull your current position for your ten most important pages, then search each query yourself. Note whether an AI Overview sits above the organic results. A high average position sitting under an AI Overview you are not cited in is worth less than the number suggests.
2. Count how many of those ten queries showed an AI Overview at all. That share, not the position number, is the one most likely to change between this check and the next.
3. Check the date of your last real look at rankings, not the date a tool last refreshed a dashboard nobody opened. If it has been longer than a week, you are checking less often than the data above says positions are moving.

Sources

The zero-click and AI Overview click-through figures come from Rand Fishkin, In 2026, Less than One Third of Google Searches Still Send a Click, SparkToro, 9 June 2026. The average position blending argument is from Dan Taylor, Is It Time To Remove Focus From Average Position In GSC?, Search Engine Journal, 17 July 2025. The ranking volatility comparison is from Yulia Deda, Google Core Update March 2026 vs December 2025, SE Ranking, 17 April 2026.

Common questions