A drop-and-pop is a short-lived ranking collapse in Google Search: a domain's keywords fall out of the results, or far down them, for a few days and then return to roughly where they stood before. The term comes from the rank-tracking industry. A tracker that reports every overnight fall as a loss causes needless alarm, and one that ignores falls misses real damage. Labelling brief, self-correcting episodes separately shows which losses matter.

The label is new, and it has not been standardised. Its most detailed public definition comes from DemandSphere, an AI search visibility platform run by Ginzamarkets, Inc. In an October 6, 2026 blog post, founder and chief executive Ray Grieselhuber described it as what happens when Google "temporarily tanks a given domain's rankings for a few days" before restoring them. The firm shortens it to DnP.

How a drop-and-pop is detected

The phenomenon cannot be seen in a single snapshot. A tracker queries Google for the same keyword daily, records the client's position and looks back across the series. DemandSphere's method rests on four building blocks.

The first is a rolling seven-day window. Each keyword's history is scanned in overlapping seven-day slices rather than calendar weeks, so an episode that straddles two weeks is not split in half. The second is a non-ranking episode: a run of consecutive days on which the URL cannot be found in the results the tracker collects. The third is a baseline, the keyword's median position before the episode. Only keywords whose pre-drop position was 20 or better, meaning the first two pages, enter the drop-and-pop charts. The fourth is imputation, the statistical filling of gaps where a day's data is missing, needed because trackers do not get a clean result every day.

From those pieces, each keyword is sorted into one of several states:

  • Stable: operationally fine once missing days are imputed.
  • Drop and pop: at least one non-ranking episode of seven days or fewer that has since resolved.
  • Tentative: ends the period not ranking, but for fewer than eight days, so the outcome is still open.
  • Eroded: still ranking, but its median position has fallen by 10 places or more.
  • Lost (recovered): a non-ranking episode of eight days or more that ended with the keyword ranking again.
  • Eroded to lost: drifted down by 10 or more positions before vanishing.
  • Lost (unresolved): still absent after eight or more consecutive days at the end of the period.

The eight-day line is the hinge. Seven days or fewer and back is a drop-and-pop; longer than that is a loss, recovered or not. The last seven days of any period are treated as a boundary zone, because imputation is weakest there, and are excluded from prevalence figures.

A worked example shows the arithmetic. A retailer's category page sits at position four for "hiking boots" through April. On a Tuesday it disappears from the first 100 results. The following Monday it is back at position five. That is a six-day non-ranking episode that resolved: one drop-and-pop. Had it returned on the ninth day, it would be filed as lost (recovered).

How common it is

DemandSphere has published numbers from one demonstration dataset, a keyword set for REI, the US outdoor retailer, which it says is not a client. Two kinds of figure appear, and they measure different things.

The first counts how many keywords had at least one episode over the whole period. Here the numbers are large: 64.8% of keywords ranking in the top three experienced a drop-and-pop, at an average of 1.6 episodes per keyword. For positions four to 10 the share was 80.6%, at 3.6 episodes, and for positions 11 to 20 it was 98.7%, at 7.4 episodes.

The second measures how many keywords were inside an episode at any given moment. That figure is small. Across all tiers, weighted by keyword count, the share was 1.8% in April 2026, peaked at 3.4% in May, fell to 2.7% in June and 2.4% in July, dropped to 0.9% in August and rose to 1.7% in September, according to DemandSphere's data as reported by PPC Land. Weighted by monthly search volume instead, the July figure was the highest, at 2.5%. Positions 11 to 20 were the noisiest, reaching 9.6% in May.

Most keywords blink out briefly at some point, but at any moment only a few per cent are missing. The lower a keyword ranks, the more often it happens.

Origin and evolution

Brief ranking swings are as old as Google. Between summer 2000 and summer 2003, Google refreshed its index roughly once a month. As each data centre loaded the new index, results differed between them for six to eight days. Webmasters called this the Google Dance, according to former Google engineer Matt Cutts, writing in December 2006. Search Engine Watch, then edited by Danny Sullivan, wrote in December 2003 of "Google Dance Syndrome" among site owners who watched positions vanish.

In summer 2003, with an update known as Fritz, Google moved to an index refreshed incrementally every day. The monthly dance gave way to smaller, constant movements that SEOs called everflux. The Florida update of November 2003, which DemandSphere's tracker lists among 173 Google updates from 2000 to 2026, showed that algorithm changes could produce far larger and more lasting shifts.

Drop-and-pop describes something narrower: a disappearance and near-complete return, measured by software. According to Grieselhuber, such episodes have occurred "to varying degrees" for years, but DemandSphere clients began noticing more of them in May 2026. That month coincided with Google's May 2026 core update, which began on May 21, and with volatility readings in mid-May that 12 tracking tools picked up while sites were dropping out of the index. DemandSphere formalised the classification in its Observability feature, launched on October 6, 2026.

Why it matters for marketers

Organic rankings feed budgets, forecasts and agency reports. A brand whose top product page vanishes on a Wednesday faces a choice by Friday: wait, rewrite the page, or shift spend to paid search. Google's own guidance cautions against reacting. Its Search Central page on debugging traffic drops, last updated in December 2025, states that "small fluctuations in position can happen at any time" and recommends avoiding radical changes to pages that already perform well.

A drop-and-pop category gives that guidance a number. If most sudden disappearances resolve within a week, a seven-day pause before intervening has empirical support. Agencies can also separate brief absences from erosion and sustained losses, which usually call for investigation.

The concept has also become entangled with a wider fight over search data. Google has steadily raised the cost of scraping its results. It removed the num=100 parameter on September 14, 2025, forcing trackers to make 10 requests instead of one. It sued SerpApi on December 19, 2025, and in August 2026 its passthrough redirects pushed rank trackers into 500 to 1,000 requests per query. Each change makes daily ranking series noisier.

Limitations and disputes

The central weakness is attribution. A tracker that cannot find a URL does not know whether Google removed it or whether the tracker was served a bad page. Today PPC Land reported that Google returns wrong results on about 11% of DemandSphere's daily rank checks, serving swapped listings to traffic it suspects is automated. Nozzle, a rival tracker, measured 8.0% of pages as poisoned. Both firms used different keyword sets and undisclosed methods, and Google has confirmed neither figure. Any of those pages could register as a drop that later "pops" when a clean result comes back.

The public evidence is also thin: one vendor, one demonstration dataset, and a retailer that is not a client. DemandSphere has not published its imputation method. No other tracker has, as of October 2026, released comparable drop-and-pop statistics, so there is nothing to check the numbers against.

Google has offered no explanation. The PPC Land report found no Google comment on drop-and-pops, and the cause - testing, index churn, spam systems or something else - remains speculation. Grieselhuber and Barry Schwartz of Search Engine Roundtable have both suggested Google's anti-scraping measures target AI companies or SerpApi rather than SEO tools. Nobody at Google has said so.

Finally, seven days, 10 positions and a top-20 filter are DemandSphere's settings, not an industry standard. Other thresholds would produce different numbers.

Not the same as

Google Dance. The monthly, data-centre-by-data-centre index refresh of 2000 to 2003. It was scheduled and index-wide; a drop-and-pop is neither.

Erosion. A gradual slide of 10 or more positions while the page keeps ranking. Erosion is slow and persistent; a drop-and-pop is sudden and reverses.

Lost (recovered). The same shape stretched over eight days or more. Duration alone decides whether an episode counts as noise or loss.

SERP volatility. Industry-wide readings from tools such as Semrush, Mozcast or Sistrix that measure how much the results move overall. Volatility is a market measure; a drop-and-pop belongs to one keyword's history.

Recent developments

DemandSphere's Observability feature adds second-order data, such as error and sampling rates, next to ranking data. Grieselhuber said it "does add a large new burden on platform vendors."

The data arrived during an unusually noisy stretch. Google's September 2026 spam update ran from September 24 to October 8, about 13 days and 16 hours. DemandSphere also recorded AI Overviews on tracked branded queries rising from 26.12% on September 1 to 82.06% on September 29. With blocking, poisoned pages and a spam update overlapping, a genuine drop-and-pop is harder than ever to tell from a measurement fault.

Timeline

  • Summer 2000 to summer 2003: Google refreshes its index about monthly; the six-to-eight-day transitions become known as the Google Dance.
  • Summer 2003: The Fritz update moves Google to incremental daily index refreshes; daily movement becomes known as everflux.
  • November 2003: The Florida update reshapes rankings for commercial queries.
  • December 23, 2006: Matt Cutts publishes his explanation of algorithm updates, data refreshes and the end of the Google Dance.
  • January 2025: Google launches SearchGuard, its anti-bot system for search results.
  • September 14, 2025: Google removes the num=100 results parameter.
  • December 19, 2025: Google sues SerpApi in the Northern District of California.
  • April 18, 2026: DemandSphere releases its Google algorithm and AI search update tracker.
  • May 2026: DemandSphere clients notice more drop-and-pops; count-weighted prevalence peaks at 3.4%.
  • May 21, 2026: Google's May 2026 core update begins.
  • July 2, 2026: Google publishes its disruption of the NetNut proxy network.
  • July 20, 2026: The court dismisses Google's DMCA claims against SerpApi.
  • August 26, 2026: Google confirms the rollout of passthrough redirects on result links.
  • September 11, 2026: DemandSphere reports the most aggressive Google blocking it has seen.
  • September 24 to October 8, 2026: Google's September 2026 spam update rolls out.
  • October 6, 2026: DemandSphere launches Observability and publishes its drop-and-pop classification.
  • October 13, 2026: PPC Land reports DemandSphere's 11% spurious-results figure and the drop-and-pop data.

Summary

Who: Rank-tracking vendors, led by DemandSphere and its founder Ray Grieselhuber, define and measure drop-and-pops; SEO teams, agencies and brands read them; Google's ranking and anti-scraping systems are the presumed source, though Google has not commented.

What: A brief disappearance of a domain's keyword rankings, lasting seven days or fewer under DemandSphere's definition, followed by a return to roughly the previous position.

When: Short ranking swings date back to the Google Dance of 2000 to 2003; DemandSphere clients noticed more drop-and-pops from May 2026, and the firm published its classification on October 6, 2026.

Where: In Google's organic search results, as observed through daily rank-tracking data, mostly for keywords in the top 20 positions.

Why: Separating short-lived drops from erosion and lasting losses helps marketers avoid overreacting to noise, though the label cannot yet distinguish genuine ranking changes from bad data served to trackers.