Two volatility spikes in six days, and Google has confirmed nothing

Volatility ran August 1 to 3 and then again into August 5 and 6, picked up broadly across tracking tools including Sistrix, Mozcast, AccuRanker, SEMrush, Algoroo, Wincher, Serpstat, DataForSEO and SimilarWeb. Google has confirmed neither round. This follows spikes on July 24 and July 18 to 19, and the last officially confirmed change was the June 2026 spam update, which ran June 14 to 26. That is roughly seven weeks of repeated, unattributed movement.

Why "unconfirmed" changes your response, not just your vocabulary

The temptation with an unconfirmed update is to treat it as a confirmed one that Google is being coy about. Sometimes that is true. Often it is not, and the difference matters operationally. A confirmed core update comes with a rollout window, a completion date, and a documented rationale. You can wait for the window to close, then measure. An unconfirmed spike gives you none of that. You do not know when it started, whether it finished, whether it was one change or four, or whether it will be reverted.

The practical consequence: the worst thing you can do this week is make ranking-motivated changes to a site. If you rewrite fifteen pages on Tuesday and rankings recover on Thursday, you have learned nothing, and you have introduced fifteen new variables into every future diagnosis.

What practitioners are reporting, and how much to trust it

Community reports during the August 5 and 6 window described traffic swings of 50 to 70% in both directions, with some sites recovering by the same evening. Others reported gains. A few observations recurred often enough to be worth noting rather than dismissing: more YouTube results on page one, multiple pages from the same high-authority domain appearing before lower-authority results, and same-day reversals.

Treat all of this as signal about what to check, not as findings. Forum reports are self-selected toward people who lost traffic, the samples are tiny, and a same-day reversal is more consistent with a test than with a ranking change. But both observations are checkable in your own data.

Half your queries now carry an AI answer

Context for both rounds: AI Overviews now appear on roughly 43% of queries according to Similarweb and about 48% according to Semrush, up from somewhere between 13 and 20% a year ago. Google has also confirmed that Gemini 3.5 Flash-Lite is one of the models Search queries get routed to, depending on the question. Google has not specifically confirmed it powers AI Overviews or AI Mode; that is an inference, and worth keeping labeled as one.

The point stands regardless of which model sits underneath. When a feature that appears on nearly half of all searches keeps changing shape, some share of what looks like ranking volatility is actually SERP composition volatility. Those require completely different fixes.

The Reddit story is the same story

In the same week, Google's Jennifer Kutz told The Verge that Reddit gets no special preference in ranking systems, and that AI features do not target specific platforms. The timing is what makes it interesting. Reddit's stock fell 12.49% after its Q2 results, with CEO Steve Huffman describing late-quarter search referral traffic as choppy and volatile, and Reddit reportedly lost visibility after the May core update and the June spam update.

The complicating fact, which PPC Land raises fairly, is the 60 million dollar per year data licensing deal signed in February 2024, which gave Google indexing rights Reddit withheld from competitors. Whether or not there is preference in the ranking systems, a commercial relationship of that shape makes the denial hard to evaluate from outside. For practitioners the takeaway is narrower and more useful than the controversy: if part of your content strategy assumes forum and UGC content occupies a permanently privileged position in results, that assumption is being tested in public. Reddit's own numbers say the position is not stable.

The diagnostic that actually works

When something moves and nothing is confirmed, the order of operations matters.

Separate ranking loss from click loss. A page holding position 3 while an AI Overview expands above it loses clicks without losing rank. Search Console shows that as flat position and falling CTR. That is not a ranking problem, and content rewrites will not fix it. We covered the measurement side of this in measuring AI-search traffic.

The first split to make. Position falling is a ranking problem. Position flat while CTR falls means the SERP changed shape above you, and a rewrite will not fix it.

Segment before you conclude. Query type, device, country, and page template. Sitewide averages hide the pattern almost every time. A drop concentrated in one template is a technical or template problem. A drop spread evenly across informational queries while commercial queries hold is a SERP feature problem.

Check the SERP itself, manually, for your top ten queries. Screenshot them. If the ten blue links are now four blue links, a video carousel, and an AI Overview, you have your answer, and it has nothing to do with your content quality.

Also worth noting this week

Search Console platform properties rolled out globally, adding Instagram, TikTok, X and YouTube content performance data. Useful for showing clients how their social content performs in Google Search, which was previously guesswork.
Google crawlers use HTTP methods beyond GET, including HEAD, OPTIONS, PUT, PATCH and DELETE, per Gary Illyes, mostly in connection with JavaScript. Worth knowing if your WAF or server config blocks non-GET methods from crawler IP ranges.

The rule for an unconfirmed week: ship planned work and technical fixes, freeze ranking-motivated rewrites, and spend the energy on diagnosis instead. If the movement was a test, you saved yourself a rewrite. If it was an update, you will fix the right thing.

What to do this week

Freeze ranking-motivated changes until the movement settles. Ship planned work and technical fixes, but do not start an emergency rewrite over a week Google has not confirmed happened.
Pull a Search Console comparison of August 1 to 9 against the prior nine days, segmented by query, device, country, and page template. Look for concentration, not sitewide averages.
Split position change from CTR change on your top 50 queries. Anything where position held and CTR fell is a SERP feature issue, and belongs in a different workstream than ranking work.
Manually screenshot the SERP for your ten highest-value queries. Build the habit now so you have a visual baseline the next time this happens.
Re-examine any strategy that assumes UGC and forum results stay favored. Reddit's own reported instability is the most concrete evidence available that the position is not fixed.

None of this is glamorous, and that is rather the point. Weeks like this reward the teams with baselines, segments, and patience, and punish the ones who rewrite first and diagnose later. Keeping that discipline on client sites is what our SEO and GEO services are for.