SEO Recap covering July 21, 2026
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)
Only 15% of ChatGPT Topic Categories Have a Clear Brand Owner, Semrush Finds
Source: Search Engine Land
- Semrush and Kevin Indig analyzed 1,094 U.S. ChatGPT categories from January to June 2026, covering more than 50,000 brands, 220,000 domains and 600,000 citations via the AI Visibility Toolkit.
- Only 15.2% of categories had a clear owner (a brand leading in at least four of five related prompts by 5 or more percentage points); 31.2% had an emerging leader and 53.7% were unsettled.
- Traditional SEO metrics predicted ownership poorly: owners led on branded search volume in 55.7% of comparisons, organic traffic in 48.4% and Authority Score in 52.5%, with only branded search volume statistically significant.
- Mentions and citations rarely aligned: only 21% of the most-cited domains in a category were also the most-mentioned brand.
- Clear owners held first place in 90.4% of month-over-month checks, while narrow leads (typical 1.3 points) were far more likely to flip than wider ones (2.9 points).
Updated SEO Rules for Bloggers: Write for Clarity in AI Search
Source: Search Engine Land
- The author updates his "write for toddlers and drunk adults" rule to include LLMs, arguing AI systems now scan, summarize and decide whether content is clear enough to retrieve or cite.
- Google's query fan-out can turn one recipe search into many subtopic queries, so bloggers should structure content around a full task, not a single keyword.
- The piece recommends using Search Console to find high-value decaying pages and genuinely improving them rather than just changing dates.
- The author urges building owned channels (email, Pinterest, YouTube, communities) since Google traffic is "rented land" that AI Overviews can erode.
Using Schema and Knowledge Graphs to Find Entity Gaps for AI Search
Source: Search Engine Land
- The author's team built a custom schema using 23 Schema.org entities plus more than 60 additional entities to map coverage gaps for university programs.
- The approach treats JSON-LD as infrastructure that explicitly asserts entities and relationships, then compares that declared model against vector embeddings of site content to surface missing entities.
- At SMX Munich 2025, Microsoft's Fabrice Canel confirmed Copilot uses schema markup, while a 2025 Search Atlas study and Mark William-Cook's research found schema doesn't drive LLM citations, a distinction the author frames as infrastructure versus a GEO hack.
- The piece recommends monitoring priority entities with prompt-tracking and brand-sentiment tools, and comparing visibility shifts against conversions and lead quality.
Category Framing Decides Which Brands LLMs Recommend
Source: Search Engine Land
- A study of 12 U.K. athletic apparel brands ran 14,140 API queries over seven days across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, testing "athleisure" versus "athletic footwear" framing.
- Changing only the category word flipped results sharply: New Balance went from 1% to 90% and lululemon from 90% to 0%.
- The authors attribute this to "category coding": the Knowledge Graph description anchors recognition, but the third-party content corpus (reviews, editorial roundups) determines recommendation.
- Recoding a KG description alone won't help, the piece argues; brands need third-party coverage in the specific category language customers use, as Nike shows by surfacing in both athleisure (77%) and footwear (90%).
- The author recommends testing five or six ways customers phrase a category query across two or three LLMs to find where your brand is missing.
Tracking AI Overviews Visibility Beyond One-Off Prompt Checks
Source: Search Engine Journal
- The session argues manual prompt tests are point-in-time snapshots, since AI engines regenerate answers each query and a confirmed citation can vanish without signal.
- Conductor's Lindsay Boyajian Hagan and Pat Reinhart outline four pillars of AI visibility: content, technical health, authority and measurement.
- The webinar covers identifying prompts worth monitoring and measuring citations, share of voice and brand sentiment over time, plus structuring headings, schema and answer formatting for extraction.
AI in Search / AI Overviews
Google Rolls Gemini 3.5 Flash-Lite Into Search
Source: Search Engine Land
- Google released Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber, with 3.5 Flash-Lite rolling out now in Google Search and the Gemini app.
- Google calls Flash-Lite its fastest, most cost-effective 3.5-class model at 350 output tokens per second per the Artificial Analysis Index, with gains in agentic workflows.
- Google cited agentic search as a use case but did not confirm whether the model also powers AI Overviews or AI Mode.
- Robby Stein said the model offers stronger instruction following and better intent understanding.
Google's Merchant Center AI Performance Pilot Adds Grouped Query Data
Source: Search Engine Journal
- Google opened a pilot showing retailers the shopping questions people ask AI Mode and AI Overviews, found in Merchant Center under Analytics, then Products, then the AI performance tab.
- The report groups questions by query type, shopping phase and product terms rather than listing individual queries, so it reveals category vocabulary, not actual searches.
- The most actionable use is checking popular attributes against gaps in your feed and working product terms into titles and descriptions; clicks and query-level data are still missing.
- Share of voice is AI impressions divided by total impressions across a fixed competitor set, and can show 0% on thin impressions or 100% with no competitors defined.
- Google plans to expand the pilot to Australia, Canada, India and New Zealand; the UK CMA has given Google nine months to add click and click-through-rate reporting for publishers.
Liz Reid Signals Google Is Building Persistent, Autonomous Search
Source: Search Engine Journal
- A Google patent, "Autonomously Providing Search Results Post-Facto, Including in Assistant Context," published February 2026 with an April continuation, describes search that stores an unanswered query and delivers the answer later when it becomes available.
- Liz Reid described information agents that keep users updated on events, exhibits or specific finance parameters without repeated checking, matching the patent's design.
- The patent lists six triggers tied to quality, authoritativeness and completeness thresholds, including when needed information does not yet exist or a resource is later updated.
- The system can deliver results proactively via notifications or unrelated assistant conversations and across multiple devices, turning search from transactional into task-based.
- Sundar Pichai has said the future of search is task-based, which the author frames as a shift SEOs need to prepare for.
Technical SEO
Audit Finds 70% of Top Retailers Miss Key Schema for Agentic Commerce
Source: Search Engine Journal
- An audit scored 141 high-traffic product pages from 29 retailers against a 10-point UCP-readiness rubric based on Google's Universal Commerce Protocol documentation.
- Basics were solid (99% carried price, availability and SKU/MPN, 96% brand), but 70% lacked all three critical fields: priceValidUntil (18%), shippingDetails.deliveryTime (13%) and merchantReturnDays (11%).
- 65% omitted a GTIN, which UCP uses to match a product across retailers for price comparison, and 15% (including Adidas UK, UGG, Converse, Christian Louboutin) blocked bots and returned 403 errors.
- Both OpenAI's Agentic Commerce Protocol and Google's UCP pull from product feeds and on-page schema, not page content, so missing fields exclude products entirely rather than ranking them lower.
- The author notes the fix is reconfiguring existing CMS fields, not replatforming, and that category pages (which outrank PDPs roughly 10 to 1 organically) carry no product schema and won't help in agentic commerce.
Organic Search & Algorithm Updates
The Dynamic Google Business Profile Playbook for AI Local Search
Source: Search Engine Journal
- The 2026 Local Search Ranking Factors report keeps primary GBP category, proximity and business-title keywords as top local-pack factors, with being open when users search now the No. 5 factor.
- A BrightLocal study of 50 businesses across 10 categories confirmed rankings tend to drop when a business shows as closed, per Joy Hawkins.
- The piece recommends operational habits: request reviews within 24 hours, respond within 48 hours, post at least weekly, and upload fresh photos at least twice a month.
- Whitespark's 2026 report added an AI Search Visibility category, with three of the top five AI visibility factors being citation and entity-based signals feeding AI Mode's local answers.
- Retailers are advised to sync top 50 products via Merchant Center with matching product schema for real-time inventory in Search and Maps.
Bing Tests an "Open" Button for Sitelinks
Source: Search Engine Roundtable
- Microsoft Bing is testing a button showing the site domain that, when clicked, reveals that domain's sitelinks on the right side, spotted by Khushal Bherwani.
- Barry Schwartz calls it an unusual sitelinks interface compared with Bing's or Google's current layouts.
Bing Tests Image Section in Shopping Panel
Source: Search Engine Roundtable
- Bing is testing an image section within the shopping results panel that sometimes appears at the top of the page, spotted by Sachin Patel.
- Barry Schwartz could not replicate it, suggesting a limited test.
Research Warns Marketing Best Practices Face "Knowledge Decay" From AI
Source: MarTech
- A paper by Prof. Rajan Varadarajan of Texas A&M in the Marketing Strategy Journal argues AI is shortening the shelf life of tactics as buyers use AI to research and shortlist vendors.
- The paper separates durable "conceptual" principles (differentiation, trust) from "instrumental" playbooks like click-earning SEO that depend on how markets currently function.
- It offers four stress-test questions about who the customer and decision-maker are, prompting marketers to check whether funnel assumptions still hold when AI agents research before a human visits.
- The author says AI differs from past shifts because it changes both marketing execution and buying behavior at once.
Published on