Insights
Weird Data and Behavior Changes
I must have read 17 articles and 97 LinkedIn posts about the removal of the &num=100 parameter and the impression drops we saw in Google Search Console.
Google confirmed the change was intentional, telling Search Engine Land, “The use of this URL parameter is not something that we formally support.”
Google hasn’t spelled out the full “why,” but the timing is hard to ignore.
Around the same time, ChatGPT’s reliance on web searches dropped significantly (Sistrix data shows it fell from over 15% to under 2.5%). Is that a coincidence?
Add to that Google hiring an Anti-Scraper Engineer, and it’s clear they’re tightening control over how their SERPs are accessed. Whether it’s about data accuracy, competitive pressure, or both, we'll feel the impact.
Bot traffic
We've spent months analyzing "The Great Decoupling," convinced that AI Overviews were tanking our organic CTRs. It seems much of that data was polluted by bot traffic from rank tracking tools.
Of course, while bot-triggered impressions exaggerated the problem, there’s still a real mix of competitive features, user behavior changes, and ongoing Google updates that affect rankings and CTR.
Even if “The Great Decoupling” was partly a misdiagnosis (impressions ≠ clicks), the fact remains that when AI Overviews appear, fewer people click through.
That has little to do with bot traffic and everything to do with actual user behavior on Google’s evolving SERPs.
The correction forces us to separate two problems: bad data from the past year, and genuine shifts in how people interact with search results.
Both matter, but for different reasons.
What do SEOs do now?
First, educate your clients immediately. Explain that their 40% impression drop isn't a traffic problem, it's a data correction.
Second, treat September 2025 as a reset point for benchmarks. Use clicks, CTR, and conversions as the primary signals.
Third, diversify beyond GSC and rank trackers. Supplement with GA4, log file analysis, and actual conversion metrics. The tools we've relied on for years are about to get more expensive and less comprehensive. Most will shift to tracking only the top 20 positions.
Google hasn't explained any of their moves, but the timing tells the story.
They're protecting their competitive moat against AI search engines while cleaning up their measurement 'glitches.'
We're just caught in the middle.
P.S. Can we please get a search console showing traffic from LLMs, AIOs, and AI Mode? Google, Open AI, anyone?
AI Mode Is Sending Almost No Traffic (Yet)
iPullRank tracked referral traffic from Google’s AI Mode (May–August 2025) to see whether it’s turning into a serious SEO channel.
The short version: it’s interesting, but it's not a volume play yet.
My brief takeaways
- Tiny referral rate. Outbound clicks from AI Mode linger below ~3%. Compared to classic Google search, it drives dramatically fewer visits.
- Click destinations skew big. When clicks do happen, they cluster on Reddit, YouTube, Wikipedia, Amazon, Facebook, and dev hubs (GitHub, Stack Overflow).
- Quick sessions, shallow depth. Sessions from AI Mode are roughly half as long as traditional Google Search, with fewer pageviews per visit.
- Queries are getting longer. From May to August, the average AI Mode query length climbed from ~7–8 words to ~10–11 words. People are treating it more like a conversational interface, which is a shift in search behavior, even if the traffic isn't following yet.
- Use is recovering, but not habitual. Session frequency dipped early, then began to rebound. That said, AI Mode has not yet formed a stable habit loop for users.
What this means for SEO strategy
- Think of AI Mode more as a credibility channel than a traffic driver. Opt for presence over volume.
- Aim to get cited. Focus your efforts on domains and formats that AI Mode already prefers.
- Build “confirm-and-go” landing pages: surface proof or the cited passage early, keep things lean, offer a clear next step.
- Track alternative metrics: citation visibility, assisted conversions, and how often your brand is referenced vs. “just clicked.”
Google's intentions are clear - now they've added AI Mode to the Chrome address bar (US).
AI Mode isn’t a flood of traffic (today), but it is shaping how people search. While you don't need to optimize for AI Mode traffic yet, positioning your brand to be cited is recommended.
Related: Google sends 831x more visitors than AI systems.
Quoteworthy
The notion that all brand discovery and selection is happening in LLMs is preposterous. It's not even happening that way for the narrow slice of the market using LLMs for search.
Yet the narrative is pushing people to put much of their focus as though it's where much of the world is and where the world will rapidly continue to go.
We are in an era where AI is part of the marketing workflow, but the old rules of marketing still apply.
—Liam Moroney, Storybook Marketing
See his entire post on LinkedIn
Real or Fake?
Pew dropped another piece of market research. This time it explored, "How Americans View AI and Its Impact on People and Society."
I pulled the image above because I suggest that "people in tech" would rate their chances of spotting AI-generated content far higher than the average American.
Only 12% of Americans are "extremely or very confident" they can detect if something is made by AI versus a person. 35% were "somewhat confident."
How do you rate yourself?
If you believe your chances are greater than 50%, try a reality check.
Reply and let me know your score because this was harder than I thought.
Do you follow me on LinkedIn? I share regular tips and stories I don't have room for here. Come and join me.
