SEO Recap covering September 9, 2026: Google's PAA turns fully AI and a new ranking factors survey

Built by Stephanie Chung·9 min read·13 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)

Jessica Bowman says AI visibility depends on operations, not just marketing
Source: Search Engine Land

  • Bowman describes LLMs as sophisticated buyer advisors that pull from customer complaints, release notes, earnings calls, vendor case studies, and employee reviews, not just marketing content.
  • She recounts ChatGPT discouraging a purchase from a retailer over product quality and return complaints, arguing "that's not something marketing can solve" and companies should treat operations as a growth area.
  • Measure AI visibility on presence and narrative rather than clicks. One client's takeaway was "we have to be comfortable with zero click," tracking where a brand wins or loses in AI narratives and sentiment versus competitors.
  • Bowman expects more SEO teams to move outside marketing as the work grows dependent on operations across departments.

ChatGPT Shopping shifts from web search to product feeds, Profound finds
Source: Search Engine Journal

  • Profound data shows feed-integrated recommendations in ChatGPT Shopping jumped from 8.26% to 61.54% on July 10, across 1,757,723 tracked prompt runs, and reached ~65% by September 3.
  • Visibility swung around the same break: of 687 tracked stores, 450 lost at least a third of their Shopping visibility comparing July 7-9 with July 10-12, while 67 gained by the same margin.
  • Retrieval concentrated on fewer stores. Top-10 reference share rose from 22.5% to 41.8% and unique merchants referenced fell over 20%. Profound estimates ~35% of July feed retrieval tied to Shopify.
  • Shopify and Etsy catalogs are already connected to ChatGPT via the Agentic Commerce Protocol. Other retailers face a waitlist for direct feed access, with a self-serve feed platform planned later this year.
  • The data covers one vendor's tracked prompts, not real shopper sessions, and doesn't show how products rank once ChatGPT pulls from feeds.

Should you drop Perplexity from your LLM tracker?
Source: Search Engine Journal

  • Siege Media CEO Ross Hudgens argued marketers should remove Perplexity from LLM trackers because its share is shrinking and it distorts aggregate visibility scores.
  • StatCounter shows Perplexity's AI chatbot referral share fell from 7.91% in June to 4.31% in August, while Gemini rose to 10.9%. Similarweb's May data put ChatGPT at 53.9% of web visits, Gemini 27.9%, Claude 9.2%, Perplexity 1.3%.
  • Google's AI search is treated as its own layer: AI Overviews reached 2.5 billion monthly users and AI Mode passed 1 billion, appearing in 43% of US searches by May 2026, up from 15% a year earlier.
  • Suggested approach is 3 tiers rather than deletion:
    • Tier 1: ChatGPT and Gemini tracked separately, plus Claude for B2B/enterprise
    • Tier 2: Google AI Overviews and AI Mode, plus Copilot for Microsoft-heavy orgs
    • Tier 3: Perplexity, Grok, DeepSeek, monitored but de-weighted, not ignored

What AI visibility scores actually measure, per Ahrefs
Source: Search Engine Journal

  • Ahrefs reports US searches for "AI search tracking" up 184% and "AI rank tracking" up 175% over the past year, signaling rising interest in these tools.
  • Mentions and citations are different signals. In 1.4 million ChatGPT prompts, Reddit URLs were retrieved at scale but cited in only 1.93% of cases, so a citation report misses much of retrieval.
  • Traditional rankings still matter: of 4 million AI Overview URLs across 863,000 keywords, 37.1% also ranked in Google's organic top 10, 26.2% ranked 11-100, and 36.7% weren't in the top 100.
  • Schema is oversold. Cited pages were ~3x more likely to include JSON-LD, but adding schema to 1,885 pages didn't significantly change citations (up 2.4% for AI Mode, up 2.2% for ChatGPT, down 4.6% for AI Overviews).
  • A visibility score can rise as clicks fall. Ahrefs found the top result's CTR was 58% lower than expected on keywords with an AI Overview, so pair visibility data with Search Console AI reports and referral analytics.

Google Merchant Center AI Performance report adds new AI search metrics
Source: Search Engine Roundtable

  • The Merchant Center AI Performance Insights report, launched in July and expanding to more regions, gained 3 new sections spotted by Brodie Clark:
    • AI Search intent: how products match customer AI searches
    • AI Search terms: popular terms to add to product descriptions for products appearing in AI searches
    • AI attributes: missing product attributes to add for AI search visibility
  • Google also revised the report's help documentation alongside the new metrics.

AI in Search / AI Overviews

Two studies on how AI is reshaping search intent
Source: Search Engine Journal

  • SparkToro's 2026 clickstream data shows 68% of US Google searches now end without a click to the open web, up from 49% in 2019.
  • Google's AI Mode has passed 1 billion monthly users. The average AI query is 3x longer than a traditional search, follow-up questions are up 40% month over month, and more than 1 in 6 searches use voice, image, or video.
  • A CUVICOM study of AI Overviews in Spanish media found 4-word queries trigger an AIO 48.1% of the time, and journalism's "6Ws" nearly guarantee one, with "why" at 92.3% and "what" at 85.7%. AIOs appear in 34.6% of evergreen searches but just 1.1% of breaking news.
  • An eye-tracking study found traditional organic links convert at 95.5% CTR and the core AIO block at 86.4%, while images, product listings, and the "Explore Further" button drew up to 100% of attention but 0% clicks.

Google's People Also Ask answers are now nearly all AI-generated
Source: Search Engine Roundtable

  • Allintitle's Also Ask Miner data shows Google's People Also Ask answers became 100% AI-generated starting August 2026, up from about 12% roughly 14 months ago.
  • AlsoAsked's sample of 19.2 million English PAA answers from 2026 put the figure at 97% in the first week of September, up from 86% in August, and on track toward 100%.

Technical SEO

Google confirms brief Search Console indexing reporting bug
Source: Search Engine Roundtable

  • For a short period on September 8, Search Console wrongly reported indexed pages, including homepages, as "Crawled - currently not indexed," with thinner or older pages most affected.
  • John Mueller confirmed it was a "short blip" resolved overnight, affecting only reporting and not actual indexing, and noted the homepage's index status is easy to double-check.

Organic Search & Algorithm Updates

Zyppy's 2026 survey ranks intent, backlinks, and content quality on top
Source: Search Engine Land

  • Zyppy's 2026 Google Ranking Factors Expert Survey polled 131 SEO professionals and collected 13,665 ratings across 103 factors, rated on a 7-point scale.
  • Respondents' picks for Google's 3 most important signals:
    • Relevance (search intent match) first at 57.1%
    • Backlinks second at 54.8%, despite Gary Illyes saying links are not a top-3 signal
    • Content quality third, with original research and first-party data rated highly
  • Links from trusted, topically relevant domains and pages with real visitors scored highest among backlink factors, while spammy backlinks were among the strongest negative signals.
  • AI-generated content rated favorably only with substantial human review; scaled AI content with little added value scored negatively. Meta descriptions ranked near the bottom.

Four risks hiding in branded search results
Source: Search Engine Land

  • Branded SERPs now mix owned and third-party results, so leaning on a few positive listings leaves a brand exposed. Build a range of trusted, authoritative assets instead.
  • AI Overviews and ChatGPT can surface negative content faster, letting one Reddit thread or critical review reach the top. Monitor movement across both traditional and AI search.
  • Look beyond blue links and AI Overviews to features like People Also Ask, autocomplete, image results, Top Stories, and Discover, which can reveal as much about a brand.
  • Diversify mentions across trusted outlets (Wikipedia, Reuters, Associated Press) to reduce dependence on any single ranking asset.
  • Run a branded search risk audit: identify branded queries, record page-one results, categorize them as positive/negative/owned/third-party, and prioritize the biggest vulnerabilities.

How to recover from Google's August 2026 spam update and manual actions
Source: Search Engine Land

  • Google confirmed the August 2026 spam update is complete, with long-standing sites seeing sharp visibility drops. A parallel wave of manual spam actions is hitting some of the same publishers.
  • Google does not target content by how it was produced. The likely issue is insufficient human oversight to hold a minimum content quality standard, not AI authorship itself.
  • Investigation steps after a drop:
    • Crawl the site and analyze server logs (tools like Lumar and JetOctopus) to see what bots hit before the change
    • Have an independent third party run the review to avoid in-house conflict of interest
    • Don't file for reconsideration until the specific flagged issue is fully fixed, since repeat rejections raise the cost
  • Getting a penalty lifted now requires full compliance with Google's spam policies, not just fixing the one flagged issue, and recovery can take months.
  • Relying on Google for 50% or more of sales is a business risk. Diversify with Bing, brand-driven direct traffic, and, for larger publishers, multiple independent brands.

The uncounted hours marketing teams spend building and fixing AI tools
Source: Search Engine Land

  • In a late-2025 METR study, 16 experienced developers ran 246 tasks half with AI. They expected AI to make them 24% faster but were 19% slower, and still believed it sped them up ~20%.
  • HubSpot reports 91% of marketing leaders say their teams use AI and 66% say their company builds its own internal AI tools, much of which never appears in project plans.
  • BetterUp Labs and Stanford surveyed 1,150 US workers on "workslop" (AI output that looks finished but isn't): 41% received it in a month, taking ~1 hour 56 minutes each to fix, over $9 million a year at a 10,000-person company.
  • Workday found companies hand back ~4 hours of rework for every 10 hours AI saves. An Upwork survey of 2,500 people found 39% of overloaded AI users cite checking and fixing output.
  • Every homebrew workflow becomes a small permanent maintenance job, and those hours usually come out of slower-ROI brand work like content, digital PR, and earning AI-cited mentions.

Mueller: sometimes there is no absolute right SEO answer
Source: Search Engine Roundtable

  • Asked whether to expand an existing article or split a sub-intent into its own connected page, John Mueller said "Sometimes there is no absolute right answer, but you can choose one, and make it the right answer."
  • The remark echoes Mueller's past lines that no site has perfect SEO and there is no perfect formula for ranking, suggesting either approach can work if executed well.
Published on
geoai searchtechnical seoorganic searchdigest