SEO & AI Search News [Week 40, 2026]

by

The week of September 30 to October 5 saw Google redraw the line on content quality.

On October 1, it told site owners to fact-check every piece of AI-generated content by hand, defined “main content” in its helpful content guide, and named fake author profiles as deception. On stage in Barcelona, Gary Illyes said scaled content is now a bigger problem than link spam.

Meanwhile, the September 2026 spam update hit a second wave on September 30 and is still rolling out.

AI Overviews spread to more than 80% of branded searches in a matter of days, a judge dismissed Penske Media’s AI Overviews lawsuit, and new reporting showed Google’s AI payments to most publishers are tiny.

On Monday, OpenAI announced visual ads inside ChatGPT image generation.

The throughline:

Google is getting ready to enforce quality harder, while AI answers move into the searches brands used to own.

Sources scanned: 300+

The week in numbers

  • 83% of U.S. desktop branded searches tracked by Ahrefs showed an AI Overview on September 29, and the September average was 73.3%, up from 61.3% in July (Ahrefs).
  • Google filters out 40 billion spam pages every day, according to Gary Illyes at Search Central Live Deep Dive Europe.
  • Most small and midsize publishers earn under 0.1% of ad revenue from Google’s AI contribution pilot (The Information, via Search Engine Roundtable).
  • 64.9% of German AI search citations point to domains with a local country or language signal (Otterly.AI, vendor data).

Subscribe Now

Search engine updates

The September 2026 spam update hit a second wave on September 30

After the first wave from September 25 to 27, a second round of ranking swings landed on Wednesday, September 30. Search Engine Roundtable logged volatility across 14 tracking tools, including Semrush Sensor, Sistrix, Mozcast and AccuRanker.

Glenn Gabe said site owners emailed him about “a huge drop right on 9/30”. He also saw Reddit’s AI-translated pages come back in Italy, Germany, Brazil, and France after losing ground in earlier 2026 updates.

As of October 5, Google’s Search Status Dashboard still lists the update with no completion date.

Why it matters:

Google said the rollout would take about two weeks, so expect more movement until roughly October 8. Keep September 24 onward annotated in your reporting, and treat drops on September 30 as part of the same update, not a separate event.

Sources: Search Engine Roundtable, Google Search Status Dashboard.

Google rewrites its content guidance: fact-check AI, define main content, no fake authors

On October 1, Google updated its guidance on AI-generated content to say: “It is critical to manually fact-check and review all AI-generated content.” Search Engine Journal reports the review now extends beyond body copy to titles, alt text, and other metadata.

The same day, Google expanded its “Creating helpful, reliable, people-first content” page with a definition of main content: “any part of the webpage that directly helps the page achieve its purpose”. That covers text and media, interactive tools, user-generated content, tabbed sections, titles and headings.

The page now calls the quality of the main content “one of the most critical factors for assessing page quality”, and judges it on effort, originality, talent or skill, and accuracy.

A new paragraph also targets fake authors: “Fabricating creator profiles (such as by using AI-generated headshots, made-up names, or false credentials to make content appear as if it was written by human experts) is a form of deception.” Similar language already existed in the Search Quality Rater Guidelines. It now addresses site owners directly.

Why it matters:

None of this is a new ranking signal. But as Marie Haynes points out, this document reflects what Google’s ranking systems are built to reward. Audit invented bylines, AI-written pages nobody checked, and pages where the main content is thin filler around ads or affiliate links.

Sources: Search Engine Roundtable, Search Engine Journal, Search Engine Roundtable, Search Engine Journal, Google Search Central, Marie Haynes.

Search Central Live Barcelona: scaled content is now a bigger problem than link spam

Google ran its first European Search Central Live Deep Dive in Barcelona from September 30 to October 2. The quote that traveled furthest came from Gary Illyes: “scaled content is now becoming more of a problem than link spam.”

He said Google filters out 40 billion spam pages a day, and Google said it is using AI to catch more. Recapping the event for We Are Roast, John Campbell wrote: “Expect more spam updates. Google is dealing with a growing flood of AI slop.”

Illyes also said that core updates target pages, not whole domains. He closed with “AI on Google is just SEO”, meaning AI Overviews and AI Mode run on the same crawling and indexing systems as classic search.

He also shared rare timing ranges, typical case first and slowest case in brackets:

  • Discovering a new URL: about 20 hours (weeks to never)
  • End-to-end indexing: about 1.5 hours (months to never)
  • Canonical change: one to three weeks (months)
  • Site move: one to three months (six months to a year or more)
  • Core update recovery: three to six months (six months to a year)

Separately, Google engineer Ariel Kroszynski showed that dark mode is coming to Search Console soon.

Why it matters:

Use the recovery range when you set expectations with clients after a core update hit. Three to six months is typical, not the worst case.

Sources: Search Engine Journal, We Are Roast, Search Engine Roundtable, Digitaloft recap via AIOSEO, Search Engine Roundtable.

AI Overviews now appear on 8 in 10 branded searches

Two datasets point the same way.

Ahrefs analyzed 232.8 million U.S. desktop search result crawls from July to September, including a panel of 100 brands from Interbrand’s Best Global Brands list. AI Overviews showed on 82.91% of branded results on September 29, and 76.2% of them sat in position one.

In a sample of 57 branded AI Overviews, Wikipedia was cited in 41, YouTube in 22, and LinkedIn in 13.

DemandSphere’s daily tracking showed AI Overviews on 26% to 35% of branded keywords for most of September. That jumped to 69.21% on September 26, 90.48% on September 27 and 82.06% on September 29. In a separate check, Chris Long of Nectiv found AI Overviews for 93 of 100 enterprise brands.

Discrepancy:

The starting points differ a lot. DemandSphere’s tracked keywords sat near 26% in early September, while Ahrefs’ panel averaged 65.1% in August. They track different keyword sets, markets and devices. Both agree on a sharp jump in the last week of September, and neither measured clicks.

Why it matters:

Your brand SERP is now usually summarised by AI before anyone reaches your site, often drawing on Wikipedia, YouTube and LinkedIn. Check what those sources say about you, and watch branded click-through in Search Console from September 26 onward.

Sources: Ahrefs, Search Engine Land.

Google’s AI payments to publishers are tiny, and Penske’s AI Overviews lawsuit is dismissed

Citing The Information, Search Engine Roundtable reported that Google’s AI contribution pilot pays about 100 publishers based on how much their content contributes to answers in AI Overviews, AI Mode and Gemini.

Most small and midsize participants receive less than a tenth of 1% of their ad revenue. One publisher earns more than $1 million a year, another $50,000 to $60,000, and some small sites made under $1,000 over several months. Participants say they cannot see how payouts are calculated.

On October 1, U.S. District Judge Amit Mehta dismissed Penske Media’s antitrust suit over AI Overviews, ruling that Penske had not shown any formal agreement with Google: “An expectation is not an agreement.” He added that the court “does not treat Plaintiffs’ alleged harms lightly.”

The same week, Cloudflare moved two paid-access products into beta. Its Monetization Gateway lets sellers charge AI agents between $0.001 and $100 per request, and Pay Per Use lets AI companies offer publishers a price for specific uses of their content.

Why it matters:

The legal route to compensation just got harder, and the voluntary one pays very little. Pay-per-crawl and licensing tools are where the leverage is for now.

Sources: Search Engine Roundtable, Search Engine Journal, Search Engine Roundtable, Press Gazette, Search Engine Journal.

Gemini adds UTM tags to outbound links, but the AI Overviews parameter was a false alarm

A Reddit user spotted Gemini appending UTM parameters to links it sends to websites, and John Mueller replied: “Nice. I see it too.” The exact values were not reported, so check your analytics for new Gemini campaign tags.

False alarm:

A “st_source=ai_overview” parameter seen on AI Overview links the same week turned out to come from a Chrome extension, not from Google. Search Engine Roundtable corrected its story.

Why it matters:

Gemini traffic should become easier to separate in GA4. AI Overview clicks still cannot be isolated from other organic clicks.

Sources: Search Engine Journal, Search Engine Roundtable.

OpenAI brings visual ads to ChatGPT image generation

On Monday, October 5, OpenAI announced a visual ad format that shows product images while ChatGPT generates images. The ads are clearly labelled and kept separate from the generated image, and testing starts later in October with select U.S. advertisers.

OpenAI also added conversion integrations with Hightouch, Tealium and LiveRamp, incrementality testing with Haus, Measured and WorkMagic, and brand safety checks with DoubleVerify and Integral Ad Science.

OpenAI says ChatGPT reaches 1.2 billion people each week, and it cites WeightWatchers reporting a 15.3% lower cost per acquisition on ChatGPT than on paid search. Both figures are company-reported.

Earlier in the week, ChatGPT added virtual try-on for clothing and accessories, plus a Favorites library for saved products.

Why it matters:

ChatGPT is building the full ad stack of a search engine (formats, measurement, brand safety) and pairing it with shopping features.

Sources: OpenAI, Search Engine Roundtable, ChatGPT release notes, Search Engine Land.

Study of the week

AI shopping answers favor big retailers, unless the user asks for independents

Tom Wells of Vaer AI ran 20,000 shopping prompts across the U.S. and Canada for Lightspeed Commerce, collecting 200,000 AI responses without web search and 260,000 with it.

Key findings:

  • Without search, national chains were recommended 63% to 70% of the time and independent retailers about 10%.
  • Shown one large and one small retailer with no size cues, the models picked the larger one 90% to 94% of the time.
  • With live search on, big retailers still took 46% to 58% of top recommendations.
  • Broad product searches gave small stores about a third of recommendations, but that fell to around 10% for named products.
  • Adding the word “independent” to the prompt raised small-store recommendations from about a third to nearly four-fifths.

Caveats:
Lightspeed sells software to independent retailers, so it has a stake in the result. Treat it as directional.

Why it matters:
For smaller retailers, generic product content will not be enough. Be specific about locality, specialism and independence, and get mentioned on the third-party sources AI systems already trust.

Source: Search Engine Land.

AI search insights

Expert review sites beat customer reviews 15 to 1 in consumer AI citations

Goodie AI analyzed 902,609 citations from 114,526 answers in ChatGPT, Perplexity, Google AI Mode and Copilot between January and August 2026.

In 21 consumer categories, expert review outlets took 4.14% of citations against 0.28% for user-review platforms, and won outright in 17 of them.

B2B software flipped: user-review platforms such as G2 and Gartner Peer Insights earned 2.25% against 1.07% for expert outlets. Amazon, the Apple App Store and Google Play each stayed under 0.5%. Vendor data.

Source: Goodie AI.

82% of commercial AI citations go to third-party pages

NP Digital analyzed 9 million AI answers across nine platforms for more than 400 enterprise brands, using Writesonic tracking data.

For commercial-intent queries, an estimated 82% of citations went to third-party sources and just 3% to the brand’s own pages.

Around 75% of answers that cited a brand’s content never named the brand. YouTube showed the strongest citation lift at 2.8x. Agency and vendor data.

Sources: Search Engine Land, NP Digital.

AI search in Germany prefers local sources

Otterly.AI tracked 3.95 million German citations from seven AI engines between August 16 and September 28.

64.9% pointed to a domain with a local signal: 51.1% to .de, .at, or .ch domains and 13.8% to localized URLs on global domains.

Reddit took 0.60% of German citations against 1.79% in the U.S. Perplexity used German sources most often (75.2%) and Claude least often (57.7%).

Why it matters:

International AI visibility needs local domains or properly localized URLs. Vendor data, but a useful reminder.

Source: Otterly.AI.

Listicles dominate AI citations in Australia, and freshness matters

Prosperity Media analyzed 122,787 citations from 467 prompts for 12 brands.

Listicles or best-of roundups appeared in 90% of prompts and took 26.9% of all citations. Cited listicles had a median age of 97 days against 136 for other pages, and 48.8% of listicle citations went to pages with “2026” in the title.

Note:

This sits in tension with this week’s Google signals. Lily Ray and others expect Google to act against self-promotional listicles built to sway AI answers.

Source: Prosperity Media.

AI agents fail on mislabelled buttons and cookie banners

MERJ’s Giacomo Zecchini ran 100 purpose-built tests plus production pages to see where browsing agents break.

When accessible labels contradicted the visual layout, one agent clicked Cancel first in 25 out of 25 runs. Cookie banners blocked interactions in about half of cases, and non-English tasks scored 9 to 18 points lower.

What to fix:

It reads like an accessibility audit: correct labels, named icon buttons, and interfaces that do not move under the agent.

Source: MERJ.

AI referrals grow, but not enough to replace organic

Brainlabs analyzed 54 client accounts covering more than 250 million sessions.

Organic traffic fell 10.5% over eight months while AI referral sessions grew 163.5%, and ChatGPT drove over 90% of AI referrals. The growth rate is high, but it starts from a far smaller base. Agency data.

Source: Brainlabs.

This week’s theme: the docs come first, the update comes next

Lily Ray’s October 4 essay argues that a major Google update is close, and the evidence lines up:

  • Google rewrote its AI content guidance, defined main content and effort, and called fake authors deception.
  • Its engineers said scaled content is now a bigger problem than link spam.
  • There have been four spam updates in 2026 against one in 2025, with the gaps shrinking from 92 to 55 to 37 days.

She points out that, historically, documentation changes like these have come shortly before big ranking updates.

Roger Montti reached a similar view in Search Engine Journal, while noting that Google has confirmed nothing.

Local businesses have their own reason to worry. Joy Hawkins of Sterling Sky says small businesses used to be largely immune to core update damage, that this immunity has disappeared, and that none of the affected sites her agency tracks has recovered since 2025.

Sources: Lily Ray, Search Engine Journal, Sterling Sky.

What to do now:

  • Find pages produced at scale with little human input.
  • Check that every author is real and verifiable.
  • Make sure the main content on money pages shows real effort and first-hand detail.
  • If you rely on paid mentions or self-promotional listicles for AI visibility, treat that as a risk, not a strategy.

On the radar

  • Gemini 4 Argon: Google previewed its new frontier model with a 1 million token output limit, up from 64,000. It goes first to trusted cyber defenders in Google’s Fairwind Program, then to paid API users and AI Ultra subscribers. No Search integration was announced. (Source: Google.)
  • Sitemaps: on Search Off the Record, John Mueller said crawl demand is “very often based on the perceived quality of a website”, so a valid sitemap does not guarantee crawling or indexing. (Source: PPC Land.)
  • Shopify: the “Allow Shopify to manage for me” setting enrolls stores in new AI shopping channels by default, now including Meta. You can turn it off under Sales channels, then Agentic. (Source: Search Engine Journal.)
  • YouTube: channels that mostly repost other creators’ Shorts with minimal changes will get less reach. No start date was given. (Source: Search Engine Journal.)
  • Microsoft Clarity: you can now customise the brand terms that split branded and non-branded queries in the AI Visibility dashboard. (Source: Microsoft Clarity.)
  • Bing: testing a “Recommended by” section in product results. (Source: Search Engine Roundtable.)
  • Google Ads: from October 19, some verified brand advertisers can use business names that differ from their destination domain. Resellers and affiliates are excluded. (Source: Search Engine Land.)

Tools worth a look

  • Quality Rater Jev (Herringbone): rates a page against Google’s Search Quality Rater Guidelines, including E-E-A-T and spam signals. Useful for the content audit described above.
  • Wikidata Knowledge Explorer (RESONEO): search entities and explore their Wikidata relationships as an interactive graph.

Worth reading

On the calendar

If you only cover three things this week

  1. Google’s content guidance rewrite: fact-check AI content by hand, use real authors, and put effort into main content.
  2. AI Overviews on more than 80% of branded searches: check what Wikipedia, YouTube and LinkedIn say about your brand.
  3. The September 2026 spam update’s second wave, with Google signalling more spam updates to come.

That is the week.

Google is telling everyone what it will reward and what it will punish, and AI answers are moving into the searches brands thought were safe. Clean up scaled content now, before the update arrives.

If you want this roundup in your inbox every week, subscribe below.

Subscribe Now

By Danny Donchev, FortuneLords. Follow on LinkedIn and X.

 

 

Leave a Comment