Google rewrites its content rules, AI Overviews take branded search

Google spent the week editing its documentation, and every edit points at scaled AI content, fake authors and gaming AI answers. People who called past big updates think one is close. Meanwhile AI Overviews spread onto branded searches, the queries brands assumed were safe. The good news is on measurement: Gemini now tags its links, and Google is testing the same for AI Overviews.

1. Google rewrote its content guidance in one week. Read it as a warning

TL;DR Google now tells sites to fact-check AI content by hand, defines "effort" as human work, and calls fabricated author profiles deception. People who track updates closely expect a big one soon.

Why it matters These are not cosmetic edits. The new "Effort" section says auto-generating pages from feeds, or producing large amounts of text with generative AI without oversight, is "little to no effort". At Search Central Live in Barcelona, Google said there are more spam updates because there is so much more new content, and that it now uses AI to catch it. The September spam update hit a second wave on September 30th, and the casualties keep looking the same: templated pages at scale.

What people are saying

• Barry Schwartz on the new line that you must "manually factcheck and review all AI-generated content", and the rewritten helpful content doc with a main content section.
• Cyrus Shepard on the fake author section Marie Haynes spotted, and his read that Google wants no scaled AI content unless you are a big brand, or Google.
• Glenn Gabe quoting the Effort definition and Google's own words: "Expect more spam updates and new spam thresholds."
• Lily Ray found 143 templated service pages, one per county, hit, noted they are the same page types companies scaled with freelancers before 2023, and predicts a massive update aimed at scaled AI content and paid mentions. Roger Montti reaches a similar conclusion.
• The other side is still loud: automated SEO agents growing new domains, "500 long tail landing pages" as a job spec, and a programmatic SEO playbook.

Do this Open your author pages. Fix any AI-made headshot or credential nobody can check. Then add a "how this was made" note to any template that uses AI.

2. AI Overviews now show on most branded searches

TL;DR Ahrefs measured AI Overviews on 83% of branded queries. Searching a brand's name no longer reliably sends people to the brand's own site.

Why it matters Branded search was the one place SEO felt safe. Now the answer box can cite Wikipedia, YouTube, reviews and competitors above your homepage, and other brands may "rank #1" in Search Console for your name. The courts will not fix it: a judge dismissed the Chegg and Penske suits, and Google's payment pilot reportedly pays some sites under 0.1% of ad revenue.

What people are saying

• Ahrefs: AI Overviews currently appear for 83% of branded searches.
• Lily Ray on what branded AIOs do to your reporting, though in her own tests the AIO sat below the paid and organic brand listings. Also: ranking #1 with 0% CTR.
• Glenn Gabe saw a surge in Gen-AI reporting when AIOs expanded in a niche.
• Kevin Indig in Time: AI Mode users write queries three times longer and behave nothing like classic searchers.
• Keval's HexClad teardown argues brands should own "is X worth it" and "X vs Y" queries, which are exactly the branded terms now getting an AIO.

Do this Search your brand plus "reviews", "pricing", "vs" and "is it worth it". Write down which sites the AI Overview cites. Those are the pages that describe you now.

3. AI traffic is starting to be measurable

TL;DR Gemini added UTM parameters to its links, Google is testing tracking parameters inside AI Overviews and AI Mode, and practitioners are building their own reporting around Search Console's gaps.

Why it matters Last week's complaint was that Search Console shows AI visibility and nothing else. Once AIO clicks carry a parameter, analytics can tell an AI Overview click from a blue-link click. With desktop CTR down in Q2, you will want that split.

What people are saying

• Roger Montti on Gemini's new UTM parameters. Barry Schwartz on the AIO and AI Mode tracking test. Matt Southern's weekly pulse has the CTR drop.
• Cyrus Shepard opted a site out of Google's AI features: AI impressions vanished in minutes and are slow to come back.
• Lily Ray asks why Generative AI impressions are such a low share of the total. Aleyda Solis shares Harry Clarkson-Bennett's fixes for the report, and Wil Reynolds proposes five new AI visibility KPIs.
• Charles Floate says rank trackers are breaking against live SERPs. Cody Schneider's answer: put Search Console and GA4 in a warehouse and point Claude Code at it. Ahrefs lists 35 Claude Code uses for SEO.

Do this In GA4, add a channel rule that catches Gemini's UTM source. When the AIO parameter rolls out, add it the same day, so your baseline starts now.

4. Gaming AI answers is getting caught

TL;DR Self-promotional listicles, automated Reddit seeding and parasite pages all worked briefly. This week's data says the models and platforms are closing the gaps.

Why it matters Across about 100 queries, AI Overviews cited 38% fewer self-promotional "best of" lists in September than in June, and left out the brand behind them 83% of the time. Otterly's 60-day Reddit experiment ended with all four subreddits banned, taking every AI citation with them. What survives is the slow work.

What people are saying

• Lily Ray's listicle numbers, and models ignoring "best SEO expert" self-recommendations.
• Glenn Gabe on Otterly's Reddit GEO experiment getting nuked.
• Linkifi ranked thin pages on Google-owned platforms within hours; Google wiped the NotebookLM ones within a day. Chris Simpson's "parasite SEO for free" via social profiles is the same bet with a longer fuse.
• The durable version: Jake Ward's list, led by the best page for each pain point and mentions on third-party sites buyers already read. Lily Ray argues query fan-out makes SEO more important, not less. A study of 412 B2B startups found message clarity only earns AIO citations if you are not already ranking.
• Aleyda Solis: local domains take 24% to 41% of AI citations outside the US, so check your market before assuming English sources win.

Do this If you publish a "best X" list that ranks yourself first, rewrite it as an honest comparison or retire it. Then put that effort into one third-party page your buyers read.

By the numbers

From SeekWinners' US shopping-search sample this week: 2,852 category terms across 12 categories.

• 60% of terms show an AI Overview. Highest: Electronics at 82%, Health and Beauty at 80%. Lowest: Furniture at 34%.
• Shopify stores get 3% to 19% of citations: 19% in Apparel, 3% in Vehicles and Parts.
• YouTube is the top cited host in all 12 categories, with 1,491 of 6,802 citations (22%). In Home and Garden it is 264 of 1,021.

Worth a click

• Gary Illyes showed how long crawling, indexing and core update recovery take.
• Crawl demand for an unknown URL comes from its parent path. Bad sections drag good pages down.
• MERJ ran 100 tests on how AI agents use websites and where they fail.
• Shopify now enrolls stores in every new AI shopping channel. Check the setting.
• Keval tripled a store's traffic with 50+ category pages and no blog.
• Cody Schneider's weekend competitor-remix workflow with an SEO agent. Read it next to theme 1.
• The whole SEO checklist in one prompt, minus the two parts you still do yourself.
• Eight months of quiet SEO: merge near-duplicates, fix every non-indexed page.
• Nicholas Dulait's GEO explainer, written to paste into Claude.

From CiteCycle

Theme 2 is the hard one to watch by hand: which sites the answer engines cite when someone asks about your brand. CiteCycle asks ChatGPT, Gemini, Claude and Google AI Overviews your customers' questions every week, separates being mentioned from being cited, and shows which third-party pages are describing you. Run the branded searches from theme 2 once by hand, then let it repeat them weekly.