SEO Recap covering September 14, 2026: Google's AI contribution pilot pays publishers

Built by Stephanie Chung·7 min read·11 stories

Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)

Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)

Two GEO experiments test what actually earns AI citations
Source: Search Engine Land

  • 2 structured experiments logged 775 citation events across up to 6 platforms (ChatGPT, Claude, Gemini, Perplexity, Google AI Mode, Grok), tracking 15 commercial-intent keywords each.
  • Earned third-party listicles dominated. In the cold-start test they drove 85.8% of source mentions versus 14.0% for the brand's own listicle, and 3 sources accounted for ~78% of all mentions.
  • Source decay was fast, with roughly half of sources dropping out of citations within 30 days, so one-time publication rarely sustains visibility.
  • Peer set matters. Removing recognized names (Lily Ray, Aleyda Solis) cut performance within days, while updating an owned listicle to include competitors lifted its mentions from 4 to 49.
  • On-page findings that helped citations:
    • Answer within the first 100 words plus a key-takeaway block near the top
    • FAQ content visible by default rather than hidden in expandable sections
    • Question-based headings and self-contained sections
    • Fresh data with visible publication dates
    • Exact-match keyword targeting with dedicated assets

Connecting AI search visibility to local leads
Source: Search Engine Journal

  • AI citation clicks account for ~1-2% of calls for CallRail's customers, a share that has roughly doubled since January, per VP Sean McCrohan at SEJ Live.
  • 2 trackable signals exist: a click on a citation in an AI response (behaves like a referral) and self-reported attribution when a customer says an assistant recommended the business.
  • AI-driven traffic skews late. About two-thirds arrives outside business hours versus half for ordinary web traffic, raising the cost of a missed call.
  • Server-side phone number swaps for AI crawlers work, but in-browser swaps do not because AI crawlers do not execute page scripts.
  • Steady Demand analysis found repeating the same local query in Gemini returns overlapping cited sources ~40% of the time and the same top business only ~7%, versus ~90% for Google's local pack.

AI in Search / AI Overviews

Google pilots paying publishers when content feeds AI answers
Sources: Search Engine Land · Search Engine Journal · Search Engine Roundtable

  • Google confirmed an "AI contribution" pilot in Search Console that pays publishers when their content contributes significantly to AI-generated responses across the Gemini app, AI Overviews, and AI Mode, first reported by Digiday.
  • Payment is tied to the generation phase where web content shapes the answer. Content that only confirms facts or gets linked after a response is generated does not qualify.
  • At least dozens of publishers have been approached, and the program has appealed more to small and mid-sized sites than large ones and widened well beyond news publishers.
  • The panel shows a monthly earnings number with limited history but no breakdown of how payouts are calculated, which one exec called "quite black box." Actual payments can differ from reported earnings due to taxes and local thresholds.
  • Publishers in the pilot said early returns are "peanuts" next to ad revenue, and one warned that accepting could weaken their leverage to negotiate better terms since Google could point to it as payment already made.

AI Overviews adds interactive knowledge quizzes
Source: Search Engine Roundtable

  • Google AI Overviews now surfaces interactive general-knowledge quizzes, extending a feature announced earlier for AI Mode.
  • The quizzes appear to build on Google's 2016 and 2021 interactive quiz tools, now embedded in AI Overviews.
  • It fits a broader pattern of Google folding existing search tools (character counters, weather, dictionary, calculators) into AI Overviews.

Technical SEO

Mueller explains why recipe thumbnails update slowly
Source: Search Engine Roundtable

  • John Mueller said it is normal for Google to take a while to refresh image thumbnails, and even longer when the old image still exists prominently on the page.
  • An Inspired Taste recipe kept showing an outdated thumbnail despite JSON-LD referencing the new image, until the owner removed the similar older image from the page.
  • The thumbnail updated shortly after, which Mueller attributed to normal reprocessing in search pipelines rather than a manual fix.

Organic Search & Algorithm Updates

September 4 ranking blip partly reverses on September 13
Source: Search Engine Roundtable

  • An unconfirmed Google ranking shift hit a number of large sites around September 3-4, showing up in Search and Discover, per Glenn Gabe.
  • Some sites that dropped sharply completely surged back on September 13, a reversal confirmed in first-party analytics, not just third-party tools.
  • Volatility trackers flag September 4 as more turbulent than surrounding days, though community chatter has stayed relatively calm.

Rethinking what content is worth when words are cheap
Source: Search Engine Land

  • Cheap AI production undercuts the agency retainer model built on charging for expensive words, which is part of why some in SEO circles demonize AI content.
  • Content is a weak force like advertising, and most keyword-driven SEO content failed to earn attention long before ChatGPT.
  • Value moves upstream from production to thinking: research direction, expertise, taste, and deciding what deserves to exist.
  • Search Engine Land data cited here shows organic search clicks fell 42% across 64 publisher sites since AI Overviews expanded, weakening the assumption that publishing a page earns a click.

AI marketing efficiency repeats programmatic's broken promise
Source: Search Engine Journal

  • Kevin Indig's argument is that AI moves marketing work rather than removing it, mirroring how programmatic converted manual buying into fraud, brand safety, and compliance labor.
  • A METR study of 16 developers on 246 tasks found they finished ~20% slower with AI while believing it sped them up ~25%.
  • Related surveys quantify the give-back:
    • BetterUp Labs and Stanford: fixing AI "workslop" takes nearly 2 hours each time, scaling past $9 million a year at a large company
    • Workday: ~4 of every 10 hours AI saves go to redoing weak output
    • Upwork: reclaimed time gets spent checking, fixing, and learning tools
  • 3 fixes: put an owner and shutdown date on every internal AI tool, track the hours spent building and maintaining tools (not just hours saved), and ring-fence slow-return work like content depth and digital PR.

Seed selection can make or break international keyword research
Source: Search Engine Land

  • The first seed phrase determines which parts of a market show up in keyword data, and a plausible translated category can return nothing at all.
  • "Commercial drone operator training" returned no data, while "drone pilot training" opened a 26,520-keyword set surfacing "FAA Part 107." Spanish showed the same pattern.
  • An August 21, 2026 test of 80 combinations across Semrush and DataForSEO found neutral seeds recovered the target "market artifact" in only about 2 of 7-8 cases, with entry failures and discovery failures as distinct problems.
  • Build a market-entry map before opening a tool, recording 4 entry points and their sources:
    • Neutral description
    • Local occupational term
    • Institutional artifact (license, card, exam, code)
    • Commercial or legacy term

Google Business Profile rolls out Collected info tab
Source: Search Engine Land

  • Google made its "Collected info" tab live, letting owners review details gathered by Google's automated assistant via call, text, or WhatsApp and delete inaccurate ones.
  • The data is used to match a business with customers, and owners can see the source and date each detail was collected.
  • Deleting collected info removes it from Google's records but does not change other details already on the profile. The feature is limited to select regions, languages, and categories.

Google tests blue "More" link in shopping results
Source: Search Engine Roundtable

  • Google is testing a blue highlight on the "More >" link inside product grids to flag when additional retailers are advertising a product.
  • Free listings normally show a less visible "& more" label when multiple retailers stock an item, which this experiment makes more prominent.
Published on
geoai searchtechnical seoorganic searchdigest