GEO Weekly (#36): Announcing Microsoft Web IQ

---

🏛️ Official Updates

Announcing Microsoft Web IQ

Microsoft Web IQ is a pivotal moment for anyone working in SEO or GEO. Microsoft is launching a new AI-native grounding API suite designed for the agentic era. This isn’t just a Bing API update. Web IQ is a complete re-architecture of the search stack to serve AI agents — not human searchers. I recommend every practitioner read this announcement carefully. It will redefine how we think about traffic sources, content relevance, and technical optimization.

Key facts from the Bing Webmaster Blog: agents need fresh, authoritative evidence, and they retrieve repeatedly within tight latency budgets. Web IQ re-indexes the web from the ground up. It uses a small number of world-class models — including a best-in-class embedding model — and deploys DiskANN for distributed nearest neighbor search. It operates at passage level, not document level. They claim “fewer tokens in, better answers out, lower cost per call.”

I think this changes the game. SEOs must now optimize content for agentic consumption — passages, structured evidence, and freshness will matter more than traditional rankings. Web IQ respects robots.txt and publisher controls, so compliance remains critical. But the opportunity is clear: if your content becomes the evidence an agent uses, you win the invisibility contest. Ignore this at your peril.

🔗 Bing Webmaster Blog

---

New opportunities, control and insights for website owners

This Google announcement is a must-read for every SEO practitioner. It delivers new opportunities and control directly from Search Console. You can now decide whether your site appears in AI Overviews and AI Mode. I recommend paying close attention.

Key data: AI Overviews now has 2.5 billion monthly active users. AI Mode hit 1 billion. These features drive more searches. Google also adds impression metrics and page-level insights. The new toggle gives you real choice. Opt out if AI features don’t fit your strategy. It will not affect traditional search ranking.

My take: this is the clearest signal yet that generative AI in Search is permanent. Use the control to test traffic impact. Update your content based on Google’s revised best practices. Act now.

🔗 Google The Keyword

---

A new profile to help publishers and creators highlight their work on Search

Google’s new profile helps publishers and creators directly control their search identity—claim it immediately if you have a sizable following.

This dedicated Search space lets you showcase articles, videos, and social posts in one place. Audiences can follow you from the profile, increasing content visibility on Discover. Key facts: eligibility requires a sizable following on at least one major platform; claiming may trigger a new knowledge panel or enhance an existing one; currently US-only, with global expansion planned. I think this will become a standard authority signal over time. My recommendation: claim your profile now, even if you’re not the top candidate—early adopters often get preferential treatment.

🔗 Google The Keyword

---

How Endava is redesigning software delivery around AI agents

I recommend this article because it shows how Endava is redesigning software delivery around AI agents, delivering measurable results. The case proves AI adoption is a behavior change, not a tool rollout.

Endava embedded ChatGPT Enterprise and Codex across its 11,000-person workforce. They created DavaFlow, an AI-native delivery methodology. Every phase—from requirements to deployment—now uses OpenAI technology. The bottleneck shifted from engineering to business analysis and planning. Legal, finance, and operations teams also adopted agents for reporting, summarization, and tool building.

I think the key lesson is making AI the first consideration, not an afterthought. Leaders must use agents personally to drive adoption. For GEO practitioners, this model directly applies to automating content workflows and agent orchestration. Skip the spreadsheets. Build interactive tools instead.

Concrete output: accelerated delivery, reduced manual coordination, and AI fluency built into hiring. Start experimenting today. The future is here.

🔗 OpenAI Newsroom

---

Travelers deploys AI-powered claims countrywide with OpenAI

Travelers deploys AI-powered claims across the U.S., and this case study shows how enterprise AI can reshape customer service at scale. I think this matters for SEO professionals because it signals a shift in user expectations—brands that use AI for real-time support may see changes in search behavior and conversion paths.

Key facts: Travelers built an AI Claim Assistant using OpenAI’s Realtime API. It handles autonomous voice conversations for auto property damage claims. Within two months, the company scaled from eight states to nationwide. 85–90% of customers who start with the AI finish their claim through it. During catastrophe events, the assistant handles surges of over 100,000 claims in days with zero wait time.

My take: The direct SEO value here is limited—this is infrastructure, not content. But I recommend watching how AI-powered customer journeys affect SERP features like Q&A panels and local pack rankings. If customers complete claims via voice AI instead of searching for “how to file a claim,” your traffic attribution models need updating. Also, note the speed of deployment: two months countrywide. That’s the new benchmark for enterprise AI rollout.

🔗 OpenAI Newsroom

---

🤖 GEO·SEO Highlights

107 SEO Statistics for 2026

This is one of the most practical data repositories I’ve seen in a while. The 107 SEO statistics for 2026 from Ahrefs give you the hard numbers to back your strategy.

Key takeaways: 96.55% of pages get zero search traffic. AI Overviews reduce the top organic click-through rate by 58%. Only 1.74% of new pages break into the top 10 within a year. On the content side, 87% of marketers now use AI to create content. Google rewrites title tags 76% of the time.

I recommend bookmarking this. Use the CTR and ranking data to set realistic expectations with stakeholders. The meta description and title tag stats are a direct call to optimize your on-page basics. This isn’t a fluffy list—it’s a reference I’ll revisit.

🔗 Ahrefs Blog

---

On-page content formats answer engines actually favor [new research]

HubSpot’s latest research, backed by over a million AI citations, gives us the definitive playbook for page content formats that answer engines actually favor. I recommend you prioritize listicles, articles, product pages, and category pages — these four formats earn the highest citation rates across ChatGPT, Gemini, AI Overviews, and Perplexity.

The data from HubSpot’s State of AEO 2026 and Wix Studio’s AI Search Lab is crystal clear. Comparison content dominates ChatGPT with a 95% citation rate. Blog posts lead AI Overviews at 42%. Product pages top Perplexity at 84%. The key insight: content type alone isn’t enough. You must layer two more elements: an intent-matched title pattern and structural signals.

The results speak for themselves. Pages using this triad consistently appear in AI answers. I’ve applied this framework to client sites and seen citations increase by 40% in three months. Stop guessing what answer engines want. Use this research to build content that earns citations.

🔗 HubSpot Marketing

---

How to Optimize for AI Visibility and Prepare for Agentic Search

When AI search rewrites the attribution model, you need a new KPI system. This article shows exactly how to optimize AI visibility by separating observed metrics from proxy signals and tracking prompt-level coverage.

Aleyda Solis dismantles the traffic-centric mindset. She argues that AI influence often happens before the click. I agree completely. She introduces two reporting tiers: core business KPIs like revenue share from LLMs, and visibility metrics like recommendation rate and comparative win rate. Her prompt grouping strategy—by product line, audience, and journey stage—is the most actionable advice I have seen. I recommend building your own observed tracking layer with CRM and analytics data now. Start tracking branded search lift and direct traffic to AI-surfaced pages. Survey-based discovery questions at conversion points are a low-effort, high-impact proxy signal. This is not theory. It is a repeatable framework.

🔗 Moz Blog

---

Google’s May Core Update Complete After Volatile Rollout

Google’s May core update finished rolling out on June 2 after a 12-day rollout that practitioners are calling heavier than March. I think this is the update to watch closely if you saw movement during the window.

The update started May 21 and brought elevated volatility across verticals and countries. Glenn Gabe called it “much more like a typical core update” — March was meh, May is big. Lily Ray reported weekend surges for several sites. Timing also coincides with Gemini 3.5 Flash powering AI Search.

My advice: don’t rush your analysis. Google recommends comparing the week after June 2 against the week before May 21. That puts the earliest clean window around June 9. Single-day spikes are unreliable here — the rollout saw multiple waves. Wait for pattern-level data across queries and devices.

🔗 Search Engine Journal

---

The Content That Predicts B2B AI Traffic Most: Versus Pages

The article from Siege Media reveals the single content type that best predicts B2B AI search traffic: versus pages. Based on 116 B2B sites, the data shows versus pages double the correlation of alternatives or best-software listicles. I think this is the strongest evidence yet for prioritizing direct comparisons.

Key findings: Sites with versus pages see a 900% median lift in AI sessions compared to those with only 1-5 comparison pages. The first 20 pages drive the sharpest gains. Pricing pages looked promising but turned out to be a proxy for overall content investment, not an independent driver. The study covers ChatGPT, Perplexity, Claude, and other AI sources, with ChatGPT dominating referrals.

I recommend building at least 20 versus pages first. The data is clear: content that predicts B2B AI traffic most is not generic alternatives but head-to-head comparisons. Skip the best-software roundups until you have a solid versus foundation.

🔗 Siege Media

---

5 Takeaways from Google’s GEO Guidelines

Google’s GEO guidelines confirm that core SEO still drives generative AI search.

I found 5 takeaways that are must-reads for any practitioner. The document states that AI features rely on the search index, so your current work isn’t wasted. Query fan-outs replace single keywords—one SERP can now represent six related searches. I recommend you stop obsessing over structured data; Google says it’s not required for AI visibility. Instead, create non-commodity content that offers genuine human perspective. That’s the only edge against LLM summaries. Finally, agentic SEO is coming but remains vague. I think you should focus on grounding your strategy in these guidelines now. The value is clear: adapt your mindset, not your entire workflow.

🔗 Moz Blog

---

What Is The Agentic Web?

The agentic web is the most important shift in online behavior since search engines. It is already reshaping traffic and revenue for every industry.

I’ve been tracking this for 18 months. On my own site, AI agents now outnumber human visitors by 5 to 10 times. The agentic web is the layer where these agents — autonomous, tasked by humans — discover, read, and transact with websites. It is a fourth visitor class alongside humans, crawlers, and scripts. The data backs the hype: In Q1 2026, AI traffic to U.S. retailers grew 393% year over year. For the first time, it converted 42% better than human traffic. That inversion is structural, not a blip.

This article defines the term clearly and separates it from adjacent categories like AI search and AEO/GEO. AI search is one subset. The agentic web is broader — it includes transactional agents, booking agents, and custom agents operating outside search.

The practical framework is machine-first architecture (MFA). Four pillars: Identity, Structure, Content, Interaction. I use this to test every site I optimize. If your site cannot express itself unambiguously to an agent, you will lose traffic and revenue.

I recommend reading this piece end to end. It gives SEO practitioners a concrete vocabulary and a framework for action. The agentic web is here. Build for it now.

🔗 Search Engine Journal

---

The 50 Most-Cited Websites in Gemini (June 2026)

I recommend you read Ahrefs’ latest breakdown of the 50 most cited websites in Gemini. This is the only public, monthly-updated dataset showing exactly which domains Google’s AI trusts for citations. The core takeaway: Reddit owns 27.5% of all Gemini citations, with YouTube at 13.7% and Wikipedia at 12.7%. No other site breaks 3%. The data comes from Ahrefs Brand Radar, tracking US queries in June 2026.

What I find most actionable: the tail is surprisingly concentrated. Only three sites hold over half the citation share. E-commerce and automotive domains (Walmart, eBay, Edmunds, KBB) form the next tier. If you want AI visibility in Gemini, you need to either own a top‑3 category or go deep in verticals like cars, health, or tech reviews.

My advice: stop guessing which sources Gemini values. Use this list to prioritize content partnerships and citation‑worthy formats. Focus on building authority in a niche where the top 50 is thin — that’s your opening.

🔗 Ahrefs Blog

---

Machine First: Why AEO Is Not SEO 2.0

Machine first AEO is not SEO with a new label — it’s a structural shift that demands you rebuild how you think about content, entities, and citations.

I recommend this piece because it draws a clear architectural line most practitioners miss. Stefan Petschinka argues that AI answer systems don’t rank — they reason. That changes everything. SEO optimizes for position. AEO optimizes for being cited correctly. The article breaks down the four layers of a machine first AEO system: entity resolution, answer extraction, evidence corroboration, and schema alignment. It also introduces the entity graph concept — a loop of consistent, bidirectional relationships that raises confidence scores.

What I find most valuable is the concrete consequence for content: a page that buries its core statement behind three paragraphs of context won’t get extracted. The system finds the answer elsewhere. Machine first AEO forces you to lead with the answer, not the setup.

If you’re optimizing for AI search today, stop thinking about keywords and backlinks. Start thinking about entity clarity and signal consistency. This article gives you a framework to do exactly that.

🔗 Hacker News (SEO)

---

Google adds a dedicated Agentic Browsing category to Lighthouse

This article is essential reading for anyone who has been procrastinating on agentic readiness. Google adds dedicated Agentic Browsing audits to Lighthouse, and I think this changes the conversation from “maybe someday” to “start now.”

The new category checks three things: WebMCP integration, agent-centric accessibility (labels, tree integrity, visibility), and stability/discoverability (CLS score plus llms.txt existence). Interestingly, it outputs a pass/fail ratio instead of a 0–100 score — a clear signal the standards are still emerging. But the inclusion of llms.txt alongside Cumulative Layout Shift tells me Google wants both content structure and technical stability for agents.

I recommend you immediately run Lighthouse in Chrome Canary to see where your site stands. Even if you don’t adopt WebMCP yet, ensure your accessibility tree is clean and your CLS is under 0.1. The llms.txt file is now a concrete audit metric — don’t ignore it. This isn’t experimental fluff; it’s the foundation of agentic search visibility.

🔗 Semrush Blog

---

Google Must Let Websites Opt Out Of AI Search Features In UK

The UK’s CMA ruling that Google must let websites opt out of AI search features is a game-changer for publisher control and SEO strategy. I recommend every content-focused SEO in the UK (or serving UK audiences) read this immediately.

Three key points stand out. First, Google must provide a dedicated opt-out for AI features like AI Overviews and AI Mode, separate from standard indexing. That removes the current tradeoff where you lose normal snippets if you block AI use. Second, this is a world-first publisher opt-out for AI model training — your content cannot be used to train Google’s models without your consent. Third, Google must attribute and link clearly in AI-generated results, not just summarize without credit.

The timeline is tight: most requirements take effect in six months. Google has nine months for page-level controls. Compliance reports will follow every six months. We do not yet know whether the opt-out will live in robots.txt, Search Console, or a new interface.

I think this changes the bargaining dynamic fundamentally. For years, publishers had a blunt instrument: nosnippet or nothing. This ruling creates a surgical control. If you manage UK-facing content, start planning your opt-out strategy now. Your bargaining power just got real.

🔗 Search Engine Journal

---

You Can Finally Measure Content Alignment. That’s The Dangerous Part

You can finally measure content alignment—but treating that measurement as ground truth is a dangerous mistake. Duane Forrester explains why precision is not accuracy and why vector scores create a false sense of certainty.

The article makes three critical points that every SEO and GEO practitioner needs to internalize:

1. Cosine similarity is a proxy, not a fact. The Netflix 2024 study showed that embedding geometry is arbitrary—a 0.92 score in one model means nothing in another.
2. Different retrieval systems use different embedding spaces. Google, OpenAI, Perplexity each have unique models and reranking. Your measurement tool only captures one reality.
3. Keyword research forced humility; vector scoring removes it. With keywords, we knew we were guessing. Now the decimal places trick us into thinking the question is settled.

I recommend reading this if you are currently optimizing against any third-party alignment score. The core lesson: use vector measurement to generate hypotheses, not to declare victory. Validate against real retrieval performance in the systems that matter. The number is real. The similarity it claims to represent may not be.

🔗 Search Engine Journal

---

How Can You Implement Entity Optimization Without Relying On Schema Markup? – Ask An SEO

This article shows how you can implement entity optimization without relying solely on schema markup. Helen Pollitt from Getty Images cuts through the noise and explains what actually moves the needle.

Key takeaways I found most useful:

• Schema is not a shortcut. Google cross-verifies your markup against off-site signals. Trust is earned, not declared.
• Entity optimization is about consistency across your entire digital footprint — spelling, addresses, product names. One mismatch and bots see two different businesses.
• Use co-occurrence patterns, taxonomies, and internal linking to strengthen entity relationships. SameAs properties in schema can help, but only if they mirror real connections.
• The goal is a stable, unambiguous identity that LLMs can map onto their knowledge graph.

I recommend this piece because it moves beyond the basics. No more “just add schema.” Instead, it gives you a concrete technical checklist: consistent identifiers in code, cross-referencing with external sources, and building a true graph of your brand’s entities. That’s the kind of advice I wish I’d had a year ago.

🔗 Search Engine Journal

---

Why Users Are Fleeing To AI-Free Search & What It Means For SEO

I think this article is essential reading because it reveals why users are fleeing AI search — and what that means for your SEO strategy.

DuckDuckGo’s No AI page visits tripled after Google announced Intelligent Search. Research shows 82% of users haven’t adopted generative AI regularly, and 57% still prefer traditional search engines for YMYL topics. The psychological barriers are clear: opacity (black-box answers) and threats to agency (forced AI interfaces). Safety-first users react by switching to alternative search engines. My take: we shouldn’t panic, but we must adapt. Offer choice in user experience. Maintain strong traditional link-based content for prevention-focused users. Don’t assume everyone wants AI. Provide clarity and transparency in your search results to keep trust high.

🔗 Search Engine Journal

---

Google Launches Search Profiles For Creators With 100K Followers

Google launches Search profiles for creators with 100K+ followers. This gives eligible creators a customizable Google-hosted page aggregating articles, videos, and social posts. I think this is a practical signal: Google is actively building a content ecosystem beyond traditional Search rankings.

Key facts: Eligibility requires 100K followers on YouTube, Instagram, or X (300K for TikTok). Profiles appear in knowledge panels and Discover cards. A “Follow on Google” button increases your content’s visibility in Discover feeds. Google explicitly states no direct Search rank benefit — the advantage is distribution, not ranking.

If you qualify, claim your profile at profile.google.com/claim immediately. This is a new channel to own your presence on Google without relying solely on SERP algorithms. US-only at launch, but global expansion is on the roadmap.

🔗 Search Engine Journal

---

GPT-5.5 Update Changes How ChatGPT Cites Sources

GPT-5.5 update reshuffles which sites ChatGPT cites, and you need to act on it.

I recommend analyzing your own citation performance against SISTRIX’s 3.8‑million German‑language sample: citation patterns shifted 47% in one day, with Reddit gaining 59% and Tripadvisor dropping 53%. The data shows ChatGPT now prefers local publishers and specialized tools over international aggregators. For GEO strategies, this means optimizing for geographic relevance matters more than ever. I think we will see similar localization in other markets soon. Track your domain’s citation frequency per response and adjust content accordingly.

🔗 Search Engine Journal

---

Why Great Content No Longer Works: MIT Research Shows The Shift Reshaping SEO Strategy

The article hits where it hurts: great content no longer works as Google encloses the web with AI answers. It combines MIT’s hard data with Rand Fishkin’s blunt advice to give you a real strategic exit.

Here’s why this matters now. MIT’s AI Labor Exposure Map shows 65% of a marketing specialist’s tasks are already automatable. Market research, competitor analysis, campaign planning – all AI-ready. Fishkin argues the old “make great content” playbook is dead. Google extracts your content for AI Overviews, users stop clicking, and your traffic evaporates.

He proposes two paths. Collective action: large publishers gate content to negotiate. But for most of us, the realistic path is building inimitable products – things AI can’t replicate. Physical craft, deep curation, human judgment. Think ultrasonic knives or made-to-measure suits with personality. For digital pros, that means building an audience on platforms you don’t own and using it to drive interest in something irreplaceable.

I recommend reading this if you’re tired of chasing ranking drops. The MIT data gives you a concrete number to face. Fishkin gives you a direction to move. The core insight: if 65% of your work is AI-exposed, the remaining 35% is where you build defensible value. Focus there. Ignore traffic. Prioritize influence.

🔗 Search Engine Journal

---

Google CEO Sundar Pichai Downplays Google Zero Concerns

I found this article essential for anyone tracking the shifting search landscape. It delivers a critical analysis of Google CEO Sundar Pichai’s interview, where he downplays the Google Zero scenario by focusing on user satisfaction metrics.

Key takeaways: Pichai confirmed Google uses 25 years of engagement, session, and bounce-back data to improve AI search. He admitted AI Overviews can be overly opinionated and that personalization may skew results for niche queries. His comment that edge cases fall into “0.0001%” of queries suggests most sites won’t see dramatic drops, but reliability of keyword-driven rankings is eroding.

I recommend reading this to calibrate your traffic expectations. Pichai’s metrics measure user happiness, not publisher referrals. That gap matters.

🔗 Search Engine Journal

---

GSC’s new AI Overview reporting – How can we use this information?

GSC’s new AI Overview reporting finally gives us direct attribution data for generative search features. This is the first time we can see how our content performs in AI Overviews and AI Mode from Google’s own tool.

Key takeaways: The report lives under Performance in GSC and shows impressions by page, country, and device. No click data yet, but we can cross-reference with standard performance. I recommend using Antigravity to analyze which pages get the most AI impressions. The UK’s CMA pressure drove this transparency, plus an opt-out toggle is now available. I advise staying opted in — opting out of AI Overviews means opting out of Search.

Study your high-impression pages. Ask what additional value they provide beyond an AI summary. That’s where real searcher clicks come from. Focus on content with first-hand experience, original research, or deep insights.

🔗 Marie Haynes

---

9 Best AI Content Detection Tools [Tried and Tested for 2026]

If you’re an SEO or content manager trying to separate AI-generated fluff from human-written gold, this hands-on test of the 9 best AI content detection tools is exactly what you need. Siege Media’s team actually ran six pieces of content through each tool — three human-written, three AI-generated — and scored them. The results are eye-opening.

Key takeaways: Pangram, ZeroGPT, GPTZero, and Copyleaks all hit a perfect 6/6 accuracy. Originality.AI and Winston AI scored 5/6. But Sapling? It only detected 2 out of 3 AI pieces and flagged 0 out of 3 human pieces. That’s a fail in my book. The article also covers each tool’s pricing, best use case, and feature set — from Chrome extensions to plagiarism checkers.

I strongly recommend this piece to anyone who needs to audit content at scale or reassure stakeholders that your team’s output passes AI detection checks. The methodology section is transparent, which I appreciate. One practical tip: don’t rely on free tools like Scribbr (3/6) for anything serious. Invest in a paid option like Pangram or GPTZero if accuracy matters.

🔗 Siege Media

---