SEO Recap covering August 3, 2026: Microsoft Clarity AI citations and Google's 304 crawl guidance

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)

Microsoft Clarity Adds Branded and Non-Branded AI Queries
Source: Search Engine Land

• Microsoft Clarity now segments the AI Citations dashboard by branded versus non-branded grounding queries, the lookups AI systems run to support a response.
• The update adds several views for separating brand-led demand from generic discovery:
  • Branded labels on individual queries in the queries card
  • A Share of Authority breakdown split by branded and non-branded queries
  • Filters to compare visibility on direct brand lookups versus broader topics
• Microsoft frames the change as a way to assess brand strength, spot discovery opportunities, and interpret citation performance shifts more confidently.

How to Measure Brand Visibility in Gemini
Source: Search Engine Land

• Gemini mentions do not appear in Search Console or Analytics, and a single query can return different responses based on follow-ups, location, personalization, and model changes, so there is no single ranking to track.
• The author recommends combining 3 approaches:
  • Manually monitor a prompt library across the buying journey, recording position, competitors, citations, and message accuracy
  • Use AI visibility tools like Profound, Scrunch AI, Otterly.AI, Peec AI, Semrush, and Ahrefs to track inclusion rate and share of voice at scale
  • Watch analytics for referral traffic, branded search growth, direct traffic, and assisted conversions
• The piece argues marketers should measure patterns (inclusion frequency, consistency, competitive positioning, trends over time) rather than react to individual responses, and review priority prompts weekly or monthly.

Combining Keyword Research and Prompt Research to Prioritize Topics
Source: Search Engine Land

• The author puts keyword volume (from Google Ads Keyword Planner cross-referenced with Semrush or Ahrefs) and prompt volume (from Profound's Prompt Volumes tool) side by side in one table for every candidate topic.
• The gap between the 2 signals sorts topics into content buckets:
  • Keyword-strong, prompt-weak topics get a classic SEO page built for the SERP
  • Prompt-strong, keyword-weak topics get extractable, citable answer content
  • Topics strong on both become flagship pillar pages built for both surfaces
• One example topic showed about 5,000 monthly searches but roughly 250,000 in prompt volume, meaning keyword data understated demand by about 50 times.
• The author advises measuring the 2 surfaces separately, splitting organic search from AI-referred traffic in reporting, and not reading an empty prompt cell as zero demand since it may roll up under a broader head term.

How to Approach GEO When You Have a Revenue Target
Source: Search Engine Land

• The author argues most GEO work optimizes for the wrong KPI, and the priority should be getting cited in the recommendation prompts that precede a purchase, not maximizing citations on broad informational queries.
• Recommended content moves:
  • Write for buyer-selection prompts with named criteria and honest tradeoffs, not generic "what is" explainers
  • Publish proprietary first-party data as a citation magnet
  • Use named authors with real credentials, and prune interchangeable content
• On technical work, the piece says to confirm AI crawlers are not blocked and pages are indexed in Bing (since ChatGPT web search relies on Bing), pull key content out of client-side JavaScript, and show visible publish and update dates. It calls llms.txt hype and schema good hygiene but not a major lever.
• The author recommends defining a money-query set of a few dozen recommendation prompts, tracking citation share only for that set, tagging AI-referred sessions into the CRM, and judging performance by the quarter given a multi-week lag between publishing and appearing in answers.

Google's Open Knowledge Format v0.2 Adds a Trust Layer No Agent Reads Yet
Source: Search Engine Journal

• The author argues OKF's value today is not machine readership (no AI agent reads bundles) but the self-audit of forcing you to declare what you know and how concepts relate.
• OKF v0.2, released July 25, adds trust metadata to each concept:
  • Source, producer, and date
  • Who verified it and whether a human or machine did
  • A staleness date and lifecycle status
• The author notes Google deliberately records raw signals rather than computing a trust score, reasoning that a score is subjective, does not port across consumers, and goes stale immediately.
• The practical takeaway is to ask 3 questions of your most important page: where the claim came from, when it stops being true, and who stands behind it.

Why AI Recommends Your Competitor and What to Do About It
Source: Search Engine Journal

• Kris Jones argues AI systems now perform the comparison step for users and hand back a decision, so brands must become the answer rather than win a spot on a results page.
• Jones estimates 70% to 80% of AI search optimization is still fundamental SEO, with the rest being third-party earned media, and advises not firing your SEO company.
• The cheapest immediate step named is pursuing credible third-party awards and best-of coverage on authoritative publications, since those are the sources models tend to cite. Publishing your own self-ranking post is called spam.
• Brand messaging consistency across the web is framed as the new NAP consistency. One example had a decade-old Reddit thread feeding tuition figures about 30% below actual prices into AI Overviews, driving confused prospects.

Reviews, Reputation, and Listings: The Local Signals AI Now Reads
Source: Search Engine Journal

• AthenaHQ's State of AI Search report found AI models cited roughly 6 to 27 domains per response (about 12 overall), with ChatGPT averaging 18.86 cited domains, Grok 26.99, and Copilot 5.77.
• Uberall's QSR benchmark found ChatGPT recommended restaurants averaging about 4.3 stars, Perplexity about 4.1, and Gemini about 3.9, though these are averages rather than thresholds.
• AthenaHQ found tracked brands' own sites were cited in only about 16% of responses, with reddit.com at 21.85% of citations and youtube.com at 10.32%, underscoring reliance on external sources.
• The recommended cleanup order is listings consistency first (name, address, phone, hours across Google Business Profile, Apple Business Connect, Bing Places), then reviews, then third-party reputation, then AI visibility tuning.

Optimizing Your B2B Brand for Autonomous AI Shoppers
Source: MarTech

• MarTechBot argues B2B buyers are delegating top-to-middle-funnel research to AI agents, shifting visibility from visual layouts toward structural readability and authority of technical product data.
• It describes how agents use products:
  • Natural-language intent filtering and synthesis that bypasses search landing pages
  • Building multi-variable comparison matrices from public technical documentation
  • Monitoring pricing and executing programmatic checkouts
  • Managing continuous replenishment based on usage patterns
• The recommended response is to prioritize structured data markup, clear semantic schemas, and high-authority independent publisher citations so inventories are readable and recommended by consumer AI agents.

---

Technical SEO

Google Recommends 304 Status Code to Conserve Crawl Budget
Source: Search Engine Journal

• Google updated its crawl budget documentation to recommend returning a 304 Not Modified response for unchanged pages so crawlers reuse the cached version and save server bandwidth.
• Google clarified that its crawl capacity limit is shared across all its crawlers, so heavy demand from one bot (such as Googlebot-Image) reduces capacity for others like Googlebot.
• The guidance targets enterprise sites over 1 million pages and medium sites over 10,000 rapidly changing pages, though it is useful for understanding crawling generally.

Google's Documentation Refresh Offers SEO Lessons on Content Updates
Source: Search Engine Journal

• Google's revised crawl budget guide adds precision and a new section, and the author frames its editorial choices as a template for improving content readability for humans and AI.
• The revisions apply 3 kinds of precision:
  • Replacing general descriptions with specific ones (for example, "5xx HTTP status codes or HTTP 429" instead of "server errors")
  • Replacing "why" explanations with the concrete effect
  • Rewriting sentences to remove clashing ideas and jargon
• The author argues removing "reading road bumps" like redundancy ("simultaneous parallel") and unnecessary anthropomorphism improves comprehension for both readers and natural language algorithms.

Bing Webmaster Tools Page and Query Level Filtering Coming to Reports
Source: Search Engine Roundtable

• Microsoft confirmed it is working on adding filtering to Bing Webmaster Tools performance reports, which currently lack any native filter box.
• The commitment came after Lily Ray asked Microsoft's Krishna Madhavan on X to add page-level filtering to queries and keywords, and he responded that they will work on it. Google Search Console already offers such filtering.

Bing Webmaster Tools SOAP/POX APIs to Be Retired August 31, 2026
Source: Search Engine Roundtable

• Microsoft will retire the legacy Bing Webmaster Tools SOAP and POX/HTTP APIs on August 31, 2026, after which requests to those endpoints will no longer be served.
• Users should migrate to the JSON/HTTP REST API, which keeps identical functionality, the same API key, and unchanged quotas, rate limits, and permissions.

---

Organic Search & Algorithm Updates

Bill Hartzer Releases Free Advanced Link Building Training
Source: Search Engine Journal

• Bill Hartzer published 50 link building lessons across 9 modules, with no signup or payment required.
• The modules cover:
  • How links work and judging link quality
  • Reading the link graph, prospecting, and outreach
  • Digital PR and earned media
  • Tactics, measurement, risk and recovery, and the future of link building
• The Digital PR module separates journalist outreach from standard link building, noting reporters do not own the site and care about the story, so the link is a byproduct of the coverage.

EU AI Act Article 50 Transparency Law Has Four Exemptions
Source: Search Engine Journal

• The EU AI Act's Article 50 is now in force, requiring labels disclosing AI-generated or manipulated content, and it applies to output accessible to consumers located in the EU regardless of where the publisher sits.
• The 4 main exemptions cover:
  • Evidently artistic, creative, satirical, or fictional deep fakes
  • Public-interest text that underwent human editorial review with editorial responsibility
  • Assistive standard editing that does not substantially alter input
  • Interactive AI where it is obvious a person is dealing with a machine
• The author notes significant ambiguity in undefined terms like "evidently artistic," what counts as human review, "public interest," "substantially," and "obvious."

Google Search Ranking Volatility Hits August 1 and Heats Up Into August 3
Source: Search Engine Roundtable

• Barry Schwartz reports an unconfirmed ranking update starting Saturday August 1 and continuing through August 3, with third-party tracking tools spiking and heavy SEO community chatter.
• Earlier unconfirmed volatility hit July 24 and July 18 to 19, following the 7-Eleven update, with the last confirmed updates being the June 2026 spam update and the May 2026 core update.
• Site owners reported result shuffling, YouTube videos getting higher preference in SERPs, and erratic Discover traffic. Schwartz notes Google is sending less traffic to publishers, prompting some large publishers to consider blocking Google entirely.

Google Search Tests Sign In to Verify Searches Are Human
Source: Search Engine Roundtable

• Google is testing a prompt asking searchers to sign in to verify they are human after clicking through several pages of results, instead of serving the usual captcha.
• The behavior was reported by Kamlesh Shukla in Chrome Incognito after browsing beyond the first few pages, though Barry Schwartz could not replicate it across browsers and searches.

Four Holiday Reports Reveal One Big Divide Over AI
Source: MarTech

• Reports from Basis, Salesforce, Attentive, and Alchemer agree shoppers are starting earlier and spending more carefully, with Attentive finding 71% plan to start before Black Friday, up 12 percentage points year over year.
• The reports diverge sharply on AI adoption depending on what they measure. Attentive says 70% of consumers use AI somewhere in the journey (Boomer adoption up from 34% to 45%), while Basis found only 17% expect to use AI while holiday shopping.
• Alchemer measured trust, finding only 35.4% of shoppers mostly or completely trust AI-generated recommendations, with reviews and friends carrying more weight.
• The takeaway for marketers is that preparing accurate product information for AI discovery is increasingly important even though few consumers are ready to let AI make purchase decisions.