Search Signal #01
Brand-owned pages take 2.8% of AI citations while review sites take 59%. Plus the August spam update, expanding AI Overviews, and what changed for B2B SaaS.
Vikas Kalwani
Founder, SAASY LINKS
01 — Being recommended is not the same as being cited
Shero Commerce tested Google AI Mode, ChatGPT and Perplexity across 60 product categories. Brand-owned pages received 2.8% of all citations. Third-party review sites took 59%. Even when a brand was recommended by name, its own pages were cited only 31% of the time.
Why it matters: this is a placement problem, not an on-page one. The pages that get cited about your category are largely not yours, so budget aimed at your own domain is aimed at 2.8% of the surface area. Read the study →
Action: Audit which third-party review and comparison sites rank for your category in ChatGPT and Perplexity. Pitch those pages for inclusion — that’s where 59% of citations live.
02 — The August spam update hit like a core update
SE Ranking analysed 100,000 US keywords across 20 industries. During the August spam update window, 16.71% of top-10 URLs fell beyond position 100 — an 82% increase in catastrophic drops versus baseline. There was also a 12% rise in new URLs entering the top 3, so this was genuine displacement.
Why it matters: a routine spam update produced core-update-scale damage, and Google’s own recovery guidance is many months rather than weeks. Read the analysis →
Action: If your rankings moved between 17–22 August, do nothing yet. Document what changed, wait 30 days, then compare. Reactive edits during volatile windows make diagnosis harder.
03 — Google is quietly auto-expanding AI Overviews
Chris Long spotted AI Overviews loading in their fully expanded state by default on some queries, pushing the ten blue links substantially further down.
Why it matters: a confirmed, silent reduction in above-the-fold organic real estate. Expect click-through to soften on affected queries with no ranking change to explain it. Read more →
Action: Check your CTR in Search Console for queries where you rank positions 1–3. If CTR dropped without a ranking change, AI Overviews likely expanded on those queries. Separate that from algorithm-driven losses.
04 — Parasite SEO penalties just got a geographic hole
Google rewrote its site reputation abuse policy. Outside the EEA, manual actions still suppress offending sections. Inside the EEA, ranking penalties no longer apply.
Why it matters: first time Google spam enforcement has openly diverged by jurisdiction. Read the policy →
Action: If you run third-party content sections (coupons, reviews, sponsored directories), audit them against Google’s four evaluation factors: visual consistency, quality difference, authorship clarity, and duplication.
05 — A judge called the AI Overviews situation “really unfair”
Judge Amit Mehta pushed back hard on Google’s defence in Penske Media’s antitrust suit over AI Overviews, noting that competing AI companies pay for content while Google obtains it free through monopoly leverage.
Why it matters: if Penske clears the motion to dismiss, the implicit bargain underneath organic search becomes a live legal question. Read more →
Action: No immediate action. But if you’re building a content moat around organic search, factor in that the legal foundation of the Google–publisher bargain may shift. Diversify traffic sources.
06 — Link carousels arrive in AI Mode
Google confirmed link carousels for developing topics are now live in AI Mode. Over 600,000 unique sources now qualify as Preferred Sources, up from 345,000 in May.
Why it matters: a rare positive publisher link surface inside AI Mode. Read more →
Action: Publish timely, well-sourced content on developing topics in your niche. These are the queries where link carousels fire. Original data and fast reporting give you the best shot at Preferred Source status.
07 — Rank Math accused of taking admin access without consent
Sybre Waaijer alleged that Rank Math v1.0.277 and later silently generates a WordPress Application Password and transmits it to Group One’s servers. Rank Math runs on over 4 million WordPress sites.
Why it matters: admin-level access without consent across one of the largest SEO plugin install bases. Worth auditing this week. Read more →
Action: If you use Rank Math, go to Users → Application Passwords in WordPress and check for any you didn’t create. Revoke unknown entries immediately.
08 — ChatGPT’s browser can now use tools that sites define
OpenAI added WebMCP support. Websites can register JavaScript functions as named tools, letting ChatGPT act on a page directly instead of simulating clicks. Page-scoped, auto-exposed, requires GPT-5.6 Sol or Terra.
Why it matters: the first credible mechanism for a site to declare how AI agents should use it. Read more →
Action: No action needed yet — the spec is experimental. But if you build SaaS tools, start thinking about which actions on your site you’d want AI agents to perform via declared tool functions rather than screen scraping.
09 — Google confirms goto URL redirects
Google moved from direct destination links to google.com/goto?url= redirects at near-100% deployment. Third-party rank tracking may be disrupted.
Why it matters: aimed at the rank-tracking industry. Watch your tools’ data through September. Read more →
Action: Cross-check your rank tracker data against Search Console through September. If your tracker shows a drop that GSC doesn’t confirm, the goto redirect is likely the cause, not an algorithm change.
10 — Shopping ads land inside AI Mode
Brodie Clark spotted shopping ads inline within the AI response itself, plus a single ad unit at the bottom.
Why it matters: ads inside the AI answer confirms AI Mode is being built as a monetised surface. Read more →
Action: If you sell a product with a shopping component, check whether your Google Shopping feed is current and your product structured data is correct. Being ad-ready when AI Mode shopping expands is better than reacting after.
11 — GEO gains decay as your competitors adopt them
“Beyond the Vacuum” reformulates GEO as competitor-aware strategy selection, showing that existing methods select rewriting strategies in isolation and the optimal strategy changes as adoption grows.
Why it matters: any fixed GEO checklist has a shelf life, and it expires fastest in B2B SaaS. Durable work is a re-measurement loop, not a static list. Read the paper →
Action: Pick your top 3 competitors. Run the same prompts monthly across ChatGPT, Perplexity, and AI Overviews. Track who gets cited and what changes. A fixed checklist won’t survive — a measurement loop will.
Also worth knowing
- Volatility kept running past the update’s official end. Rankings spiked across an unusually broad set of trackers after 21 August.
- Google’s spam team is still well resourced. Rajan Patel, VP of Engineering for Search, confirmed on X.
- The spam update targets mass-generated content, not AI content generally. Google published research on identifying networks of AI-generated spam.
- John Mueller says there is nothing special to do for GEO. Worth citing, but note it covers Google’s own surfaces only.
- On the calendar: 15 September is when Cloudflare blocks training crawlers by default on ad-supported pages for new domains.
SAASY LINKS
We build the visibility this newsletter reports on
Editorial links, AI citations, and organic pipeline for B2B SaaS. Senior-led, 90-day sprints, no long-term contracts.
Book a Call