SEO Recap covering July 23, 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)
Yelp licenses reviews and ratings to OpenAI for ChatGPT local answers
Source: Search Engine Land
- Yelp is licensing reviews, ratings, photos, and business details to OpenAI, so ChatGPT can surface that content in local queries with Yelp branding and links, though OpenAI controls the presentation.
- Yelp's Request a Quote feature will work inside ChatGPT, letting users contact local service providers for quotes directly. The deal does not stop Yelp from licensing to other AI companies, and terms were undisclosed.
- Morning Consult found 65% of Americans have used AI search but only 15% trust it "a lot," and 72% think AI platforms should always identify their sources.
- Yelp is separately suing Google over local search self-preferencing, after an earlier licensing relationship broke down.
seoClarity webinar: proving AI search wins with split tests
Source: Search Engine Journal
- Adding FAQ sections to test pages lifted AI citations across roughly 1,000 measured prompts, and removing them dropped citations back down, which the team calls proof of causation rather than correlation.
- On June 3 Google launched Search Console reports for AI Overviews and AI Mode showing per-URL appearance in AI features, but they cover impressions only and ChatGPT, Claude, and Perplexity still need third-party tracking.
- The recommended method builds a golden set of prompts across the funnel, tiers them by where the brand almost wins, then uses a correlated control group with fixed baseline and test windows instead of live A/B splits.
- Meta description and listicle formatting tests failed, and presenters stressed that well-optimized, technically healthy sites still perform best in AI search, with AEO as a layer on top.
ChatGPT ships its own "not provided": attribution arrives through the ads door
Source: Search Engine Journal
- The author argues platforms will build closed-loop attribution for paying advertisers, not for organic practitioners, drawing a parallel to Google's 2011 move that hid organic keyword data as "not provided."
- OpenAI already offers advertisers a server-to-server conversions pipeline with a pixel, events API, standard events, deduplication, and a privacy-preserving identifier, while organic operators get only robots.txt toggles like OAI-SearchBot and GPTBot.
- He splits "attribution" into 3 problems and says 2 are already solvable without the platforms:
- Referral attribution, measurable now via referrers, UTMs, and a classification rule treated as a floor
- Incrementality, provable yourself with geo holdouts or on-off tests
- Influence in the dark funnel, hard, handled with self-reported attribution and branded-demand correlation
- Cited figures diverge sharply: Microsoft Clarity found AI-referred visitors converting to signups at 11 times the rate of organic search, while Marketing Science research across 973 sites and $20 billion in revenue found LLM traffic converting below every traditional channel except paid social.
- Before buying a vendor, he says ask what data source closes the loop: first-party sources are legitimate but bounded, and any claim of signal from inside OpenAI or Anthropic is misrepresentation.
Semrush: ChatGPT has no clear brand leader in most categories
Source: MarTech
- Across 1,094 US categories tested with 5 buyer prompts each (definitions, comparisons, alternatives, use cases, purchase decisions), only 15.2% had a clear brand leader. 31.2% had an emerging leader and 53.7% had none.
- To qualify as owner a brand had to appear in at least 4 of 5 responses and lead the next competitor by 5 points. In the highest demand topics, which drove 98% of AI search volume, only 11.3% had a clear owner.
- Branded search volume was the only traditional metric aligned with category ownership. Organic traffic and Authority Score showed little relationship, and only 21% of the most cited domains were also the most mentioned brand.
- Established leaders held first place in over 90% of month-over-month comparisons, while leadership in unsettled categories changed nearly 2,000 times during the study.
AI in Search / AI Overviews
Google says AI Max opens billions of unmonetized searches to ads
Source: Search Engine Land
- Google said AI Max is now out of beta with more than 500,000 advertisers, and that AI Max or Performance Max users see an average 15% lift in conversions or conversion value at similar return on ad spend.
- Gemini has improved Shopping ad relevance for complex queries by about 20%, letting Google match longer conversational searches to ads without relying on exact keywords.
- Google detailed new AI Mode ad formats:
- Highlighted Answers, labeled sponsored links placed inside AI-generated list responses
- Contextual sitelinks generated from conversations
- Direct Offers surfacing promotions during planning, with IHG Hotels named as a partner
- The company framed AI Max as a way to create new ad inventory inside Search, though advertisers get less visibility into how queries are matched.
Technical SEO
How to decide which technical SEO debt to fix, monitor, or ignore
Source: Search Engine Land
- The author argues crawlers surface tens of thousands of issues sorted by what is easy to count, not by impact, so the real work is prioritizing by business impact and fix effort rather than chasing a clean crawl report.
- He sorts issues into 4 action buckets: fix now, fix soon, monitor, and ignore for now, mapping each to whether the issue touches important pages, templates, or revenue paths.
- Each issue is scored across 5 factors, then translated into P0 to P4 priority levels:
- SEO impact on crawling, indexing, rankings, or traffic
- Business impact on leads, revenue, or signups
- Scale (one page, one template, or thousands)
- Risk of compounding or future problems
- Effort in dev, content, QA, or stakeholder work
- The tactical core is segmenting a Screaming Frog crawl by section, template, or intent, then layering Google Search Console, GA4, backlink, rank tracking, log file, and CRM data so the right priority becomes clear.
AI browsers are a workaround for a web that lost its semantics
Source: Search Engine Journal
- OpenAI retired its Atlas browser on July 9, 2026, folding it into the ChatGPT desktop app and a Chrome extension only 9 months after launch, following the shutdown of its Sora video app.
- The author argues AI agents read the accessibility tree and document structure, not pixels, so a styled div that behaves like a button is invisible to both agents and screen readers, and a watchable browser adds nothing for the agent.
- Vision-based "computer use" agents that operate a rendered page work on any site with no owner effort, but the author calls this a per-visit tax that stays slow and fragile because nothing underneath gets fixed.
- His fix is putting the fundamentals back: consistent messaging, fast readable pages without a JavaScript wall, and markup a machine can identify and use, so the site is ready for any agent in any shell.
Organic Search & Algorithm Updates
EU fines Google 890 million euros under the DMA for Search and Play
Source: Search Engine Land
- The European Commission fined Google 460 million euros for self-preferencing its own shopping, hotels, transport, and sports results over third parties in Search, and 430 million euros over Google Play anti-steering restrictions, roughly $1 billion combined.
- Google must treat third-party services fairly and non-discriminatorily in search results, and let Play developers freely communicate and conclude deals with users outside the store.
- Google has 60 days to comply or risk periodic penalty payments of up to 5% of worldwide turnover.
- The author expects Google to appeal and to make further Search changes, especially in Europe.
Why being competent is not enough to get hired in SEO anymore
Source: Search Engine Land
- The author argues that with hundreds of qualified applicants per SEO role, years of experience and standard skills are now the baseline for an interview, not the reason for an offer.
- He recommends building tangible proof of curiosity and initiative, such as a small SEO tool, a testing site, a newsletter, a GitHub project, a case study, a Looker Studio dashboard, or nonprofit volunteer work.
- The point of such projects is not popularity or revenue but the stories they give you in interviews, which stick with hiring managers longer than another discussion of keyword research or audits.
Buying backlinks for AI visibility risks repeating Penguin-era penalties
Source: Search Engine Journal
- The author warns that services repackaging old link-building tactics as AI visibility solutions ignore why Google's Penguin algorithm drove those practices underground in the first place.
- He says he has seen more link-based penalties in the past year than in the prior 5, driven by people buying backlinks for AI purposes without grasping how the manipulation appears in Google and Bing indexes.
- His argument is that doing something for AI does not make it an agnostic SEO practice, and that Google's inevitable cleanup of this spam will leave high-quality, unpenalized sites more valuable.
Alphabet's Q2 shows record revenue and negative free cash flow
Source: Search Engine Journal
- Q2 2026 revenue was $119.8 billion, up 24% year over year, with Search & Other at $63.3 billion, Google Cloud $24.8 billion, subscriptions and devices $12.9 billion, and YouTube ads $11.1 billion.
- Alphabet reported negative free cash flow of $5.855 billion, with $39.069 billion operating cash flow against $44.924 billion in capital expenditures, roughly double the prior year's property and equipment spend.
- CapEx is expected to reach $180 to $190 billion this year, about 6 times 2022 levels, with the CFO saying 60% went to servers and 40% to data centers.
- Alphabet raised $49.6 billion in an equity offering to help finance AI infrastructure, signaling operations alone are not covering the spend.
Pichai says Google needs Gemini 4 to compete at the frontier
Source: Search Engine Journal
- On Alphabet's Q2 call, CEO Sundar Pichai said Google needs to improve coding and agentic coding and that a larger Gemini 4 base model, now in pretraining, is necessary to stay competitive.
- The unreleased Gemini 3.5 Pro remains delayed over coding issues, now "testing with partners," after a late June training data update fell short.
- Google shipped Gemini 3.6 Flash, which it says produces 17% fewer output tokens than 3.5 Flash and scored 49% on the DeepSWE coding benchmark versus 37% for 3.5 Flash. A cheaper 3.5 Flash-Lite tier is being integrated into Google Search.
- In June, Gemini co-lead Noam Shazeer and AlphaFold's John Jumper left for OpenAI and Anthropic amid concerns about Google's position in AI coding.
Google Search revenue growth eases after a year of acceleration
Source: Search Engine Journal
- Google Search & Other revenue rose 17% year over year to $63.27 billion, the first slowdown after four straight quarters of acceleration that peaked at 19% in Q1 2026.
- Total Alphabet revenue hit $119.8 billion, up 24%, with the Gemini app now at 950 million monthly active users and models processing 22 billion API tokens per minute.
- Pichai tied Search growth to AI features driving query growth, and Schindler credited retail, finance, technology, and media verticals plus AI Max targeting hard-to-monetize searches.
- The new Search Console generative AI report shows impressions in AI Overviews and AI Mode only, not clicks or CTR, so revenue strength does not indicate site traffic.
Large publishers weigh blocking Google Search entirely as clicks fall
Source: Search Engine Roundtable
- A Wall Street Journal report says publishers including USA Today, Politico, Reuters, The Economist, People, and possibly Reddit are considering disallowing Google's crawler as clicks keep declining and the value exchange no longer pays off.
- USA Today's CEO said "enough is enough" and People Inc.'s CEO said blocking Google is "100% on the table," while The Economist said full exit is not currently on its table because "traffic is still traffic."
- The WSJ chart of publisher declines uses Semrush data, and the moves follow reporting on how AI search imperils the open web, plus Reddit's $60 million Google content deal coming up for renewal after recent core and spam updates.
EU fines Google 890 million euros for DMA breaches on Search and Play
Source: Search Engine Roundtable
- The European Commission issued a 460 million euro fine for self-preferencing Google's own Shopping, Hotels, Transport, and Sports results, and a 430 million euro fine for Play anti-steering restrictions, about $1 billion total.
- Google has 60 days to implement non-discriminatory rankings and let developers direct users off-platform, or face periodic penalty payments of up to 5% of global turnover.
- Google has already begun testing changes to search display, shopping ads, and Play steering terms, and is discussing how the principles apply to AI Overviews and AI Mode, with EU hotel results already changing.
Published on