Bing just dropped a game-changer for anyone tracking new AI visibility<\/strong>: four new metrics\u2014Intents, Topics, Citation Share, and Compare\u2014are now in preview inside Bing Webmaster Tools. This is official data from Microsoft, not a third-party estimate. I recommend every SEO and GEO practitioner dive into this immediately.<\/p>\n What makes this update different? First, the Intents<\/strong> feature classifies grounding queries into categories like Informational, Commercial, or Research. Instead of just seeing that<\/em> you were cited, you now understand why<\/em>. For example, an e-commerce site can discover it gets cited mostly in shopping-oriented AI answers. That directly informs content strategy.<\/p>\n Second, Topics<\/strong> group related queries into thematic clusters (“Solar Energy” instead of “solar panels” + “solar efficiency”). This mirrors how AI models reason\u2014across concepts, not keywords. It helps you spot gaps in coverage or emerging authority areas.<\/p>\n Third, Citation Share<\/strong> shows your site’s percentage of all citations for a given query. Total citation count alone is noisy; this ratio tells you how much of the AI “mindshare” you actually own. It’s a directional metric, not a ranking\u2014use it to track your relative presence over time.<\/p>\n Finally, Compare<\/strong> lets you overlay two date ranges to spot shifts from content updates, seasonality, or model refreshes.<\/p>\n I see this as a direct signal that branded search is being reshaped by AI. Don’t wait for third-party tools; start poking around in Bing Webmaster Tools today. Connect these insights to your own content performance data and adjust your GEO strategy accordingly.<\/p>\n \ud83d\udd17 Bing Webmaster Blog<\/a><\/p>\n OpenAI’s latest work on improving health intelligence in ChatGPT is crucial for GEO practitioners.<\/p>\n I recommend examining their physician-led evaluation approach. Over 230 million people use ChatGPT for health each week. GPT-5.5 Instant reduced factuality issues by 71% on production traffic. It now performs on par with frontier Thinking models on HealthBench Professional. Physicians rated its responses higher than human-written ones across 3,500 reviews. Key improvements include fewer missing red flags and better context handling. The article shows concrete response comparisons, like explaining MRI necessity for sciatica. Use these insights to optimize health content for AI-driven answers. Focus on accuracy, clear communication, and appropriate escalation.<\/p>\n \ud83d\udd17 OpenAI Newsroom<\/a><\/p>\n This new OpenAI Academy gives SEO and GEO teams a structured path to build repeatable AI workflows. I think it’s a practical resource for anyone scaling AI adoption in content or search work.<\/p>\n Three courses cover the full spectrum: AI Foundations teaches prompting and output review; Applied AI Foundations turns those into repeatable workflows; Agents and Workflows shows how to direct agent-assisted tasks with human oversight. Certificates give teams a way to track progress. OpenAI partnered with BCG, Accenture, and BBVA to shape the content.<\/p>\n The article lacks specific SEO use cases, but the workflow design directly applies to automating research, drafting, and quality checks. I recommend agencies and in-house teams use these courses as a shared baseline before building custom GEO pipelines. The curriculum evolves with OpenAI’s models, which keeps it relevant for search practitioners.<\/p>\n \ud83d\udd17 OpenAI Newsroom<\/a><\/p>\n I recommend this article because it proves something we all suspect but rarely test cleanly: AI without live Google data is guessing.<\/strong> Adam Ga\u0142\u0119cki ran a controlled experiment where the same model competed against itself \u2014 the only difference was access to real-time SERP data via an API. The results are brutal.<\/p>\n Key takeaways: First, keyword research from the naked model had only a 42% verification rate against live data. It hallucinated plausible searches that nobody actually types. Second, when predicting SERP features and domains, the model missed AI Overviews and Perspectives entirely, and invented 6 out of 15 competitors. Third, the live data revealed a critical homonym issue \u2014 \u201cSERP\u201d also means \u201cSupplemental Executive Retirement Plan,\u201d which the naked model completely ignored. The one win was query fan-out at 80% accuracy, but even then you can\u2019t tell which 20% are lies without checking.<\/p>\n My take: this is why I always pair AI outputs with real-time SERP verification. The model is good at patterns, but terrible at ground truth. Use live data as your answer key, not a supplement.<\/p>\n \ud83d\udd17 Hacker News (SEO)<\/a><\/p>\n Google\u2019s updated guidance finally gives us the green light on LLMs.txt for AI SEO. I think this is a significant shift from the previous discouraging tone. Google now explicitly states using LLMs.txt is \u201ccompletely fine\u201d for other AI services and systems. It won\u2019t help or hurt your visibility in Google Search itself. But that\u2019s the key insight: Google still ignores these files entirely.<\/p>\n What changed? The original guidance implied special markup wasn\u2019t needed at all for generative AI search. The update narrows that statement to \u201cGoogle Search (including its generative AI capabilities).\u201d This leaves the door open for optimizing for ChatGPT, Perplexity, and other AI surfaces.<\/p>\n My recommendation: definitely create and maintain LLMs.txt files if you want to be cited by non-Google AI tools. Just don\u2019t expect any ranking boost in Google. This is a smart, practical update from Google. It acknowledges the multi-platform reality we now operate in.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n I recommend building your AI search strategy around category entry points (CEPs). These are the real-life situations that trigger a buyer to consider your category. My experiment at Semrush proved CEP-anchored content works. One article earned a citation every week for four months. Another lifted our share of voice from 15% to 26% in one week.<\/p>\n Why does this matter? AI search lets users describe their full situation. CEPs map directly to those prompts. One CEP captures many different phrasings. Mental availability becomes measurable through brand mentions and AI citations.<\/p>\n My advice: Identify your buyers’ category entry points first. Write content that matches those specific moments. Then watch how often AI systems cite your work. This approach beats generic keyword targeting every time.<\/p>\n \ud83d\udd17 Semrush Blog<\/a><\/p>\n What your CMO needs to know<\/strong> about AI search is that GEO measurement looks nothing like traditional keyword ranking. Tom Critchlow makes this crystal clear in his conversation with Ross Hudgens.<\/p>\n I think the biggest takeaway is this: GEO behaves like brand measurement, not SEO. You stop tracking prompts and start tracking share of model, brand awareness, and how LLMs describe you. Tactical prompt monitoring won’t move the needle.<\/p>\n Three points I’d highlight to any leader:<\/p>\n I recommend sharing this with your CMO. It reframes the conversation from “are we ranking?” to “are we mentioned and how?” That’s the real battle in 2026.<\/p>\n \ud83d\udd17 Siege Media<\/a><\/p>\n Calling yourself best in a self-promotional listicle is now helping your competitors win in AI search\u2014and that’s backed by original 2026 data. This article by Lily Ray is a must-read for any SEO practitioner who still thinks ranking No. 1 in your own article is a safe bet for AI visibility.<\/p>\n The study tracked 100 B2B “best [category]” queries across Google’s AI Overviews on three dates. Here’s the kicker: 69% of the time<\/strong> the self-promoting brand got cited as a source but was left out of the actual recommendation. Google instead recommended the competitors listed in that same article. The recommendation is what drives real value\u2014citations alone generate near-zero clicks (just 1% in a Pew study). I believe Google has effectively decoupled citations from recommendations, making self-promotional listicles a liability for most sites.<\/p>\n My take: If you’re an established brand, publishing these pages still might work, but it erodes trust and invites future penalties. If you’re a smaller brand, stop. You’re literally campaigning for your rivals. Focus on earning third-party recommendations instead. This is the most actionable AI search research I’ve seen this year.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n If you run a small merchant site, stop searching for protocol specs. This article gives you the real checklist for agentic commerce small merchant readiness. Your platform already decides which spec matters.<\/p>\n I recommend reading this because it cuts through the technical noise. The author confirms that Shopify, BigCommerce, Wix, Squarespace, WooCommerce, Stripe, and PayPal all have agentic commerce integrations live or rolling out. Your only job is configuring the admin panel and cleaning product data.<\/p>\n Key takeaways:<\/p>\n I think the most valuable insight here is that clean product data accounts for 90% of the work. The protocol layer is already handled by your platform. Focus on descriptive titles, complete descriptions, accurate pricing, and consistent categorization. That is where agents succeed or fail on your storefront.<\/p>\n The author also names a 90-day path to a fully agent-ready storefront without writing code. For small merchants without engineering teams, this is the only playbook you need.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n Publishing more content is now weakening your SEO.<\/p>\n I recommend you stop treating volume as a strategy. In 2026, AI retrieval systems penalize semantic dilution. They reward clarity, consolidation, and authority density. Multiple overlapping pages fragment embeddings. Your own pages compete for retrieval dominance. No single fragment wins. I see large sites with decent traditional rankings vanish from AI summaries. The fix: consolidate duplicated content, build one authoritative page per topic. Adopt entity-based structuring and schema markup. Publish less, but make each page the definitive answer. That\u2019s how you earn citations in ChatGPT and Google AI Overviews.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n Agent A finally makes good on the “AI for SEO” promise. This article gives you 11 ways to automate<\/strong> tedious SEO tasks using Ahrefs’ own marketing agent. I recommend reading it whether you use Agent A or not.<\/p>\n The piece delivers concrete, copy-paste workflows. For example, Sam (Ahrefs VP of Marketing) built a keyword research tool that transforms a niche like “coffee” into graded keyword clusters, complete with content briefs, in 20 minutes. Another workflow turns every weekly site audit into a GitHub PR overnight. A third catches decaying posts before they flatline, comparing traffic quarter-over-quarter.<\/p>\n Each of the 11 use cases includes a starter prompt you can drop into Agent A’s chat. That is the article’s real value. You do not need to be a developer. The prompts are short and actionable.<\/p>\n The article also shows how Agent A connects to Slack, GitHub, Notion, Linear, WordPress, and more. That makes these automations practical for real teams.<\/p>\n My recommendation: start with the weekly site audit to GitHub PR workflow. It fixes the classic problem of audit findings sitting untouched. The PR lands Monday morning with a checklist and audit data attached. Your developers will thank you.<\/p>\n \ud83d\udd17 Ahrefs Blog<\/a><\/p>\n Agentic marketing is the real deal for moving from strategy to execution in one loop. I recommend this Ahrefs piece because it gives you the cost data, agent taxonomy, and 11 copy-paste prompts to start today.<\/p>\n Search volume for \u201cagentic AI\u201d jumped 84\u00d7 in three years to 122,175 monthly US searches. LLM costs dropped 15\u201320\u00d7 since 2023, so a multi-step agent run now costs cents. Ahrefs\u2019 own hackathon produced 16 internal apps, zero code from the team. The article explains three agent types, key terms like skills and MCP, and specific SEO use cases including AI mention gap analysis.<\/p>\n The primary keyword \u201cagentic marketing\u201d gets grounded in real tools. I suggest reading it, then testing one prompt this week. The starter prompts are battle-ready, not theory.<\/p>\n \ud83d\udd17 Ahrefs Blog<\/a><\/p>\n This article delivers the most actionable data I’ve seen on the 10 SEO trends<\/strong> reshaping our industry in 2026. Ryan Law at Ahrefs backs every trend with real search volume data, not speculation.<\/p>\n The key trend is query fan-out, search volume up +2,550% YoY. It changes what we optimize for. AI search engines like ChatGPT generate hidden queries behind every prompt. The winning move is topical clusters, not chasing specific fan-out queries.<\/p>\n AI Overviews now reduce clicks to the #1 result by 58%. Google keeps those clicks. AI Mode adds another layer of zero-click traffic loss. The unit economics of SEO have changed permanently.<\/p>\n I was most struck by the AI content section. The old trade-off (faster content, worse quality) is gone. Content engineering with agentic harnesses and tool calling now produces genuinely good content. The data shows steadily climbing interest in AI creation.<\/p>\n I recommend this article to anyone managing SEO budgets in 2026. It forces hard decisions about where to invest: deeper topical authority, AI-ready infrastructure, and branded search. The old top-of-funnel playbook is failing. This article shows the new path forward with cold, hard numbers.<\/p>\n \ud83d\udd17 Ahrefs Blog<\/a><\/p>\n The integrated search brief is the single most effective way to align SEO, PPC, and content teams in the AI search era. This framework forces teams to start with a business outcome, not a keyword or ad group.<\/p>\n Corey Morris illustrates this with a real B2B construction firm example. The brief covers three pillars: business objective, audience and search intent, SERP landscape. It accounts for AI Overviews and AI Mode, which change user behavior. Without a connected plan, each team pursues separate priorities. That wastes resources and misses opportunities.<\/p>\n I recommend adopting this shared operating agreement immediately. Use a table to define channel roles, intent types, and measurement KPIs before anyone creates content. Silos no longer work in today’s fragmented search results.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n We analyzed 137K domains and found 97% of llms.txt files never get fetched.<\/strong> That\u2019s the bottom line from Ahrefs’ massive server-log study. The SEO industry has been chasing a phantom. Here are the facts you need.<\/p>\n I think this study kills the \u201cllms.txt for AI visibility\u201d narrative cold. The file is a token-saver for coding bots, not a ranking signal. If you\u2019re spending time crafting one for SEO purposes, you\u2019re better off auditing your server logs to see what actually hits your site. John Mueller was right: you\u2019re getting very little AI agent traffic today. Focus on fundamentals \u2014 good content, clear structure, fast pages. That\u2019s what real bots and users both reward.<\/p>\n \ud83d\udd17 Ahrefs Blog<\/a><\/p>\n This is the most actionable guide I\u2019ve seen on how to optimize agentic web<\/strong> presence. Tushar Pol from Semrush lays out a five-layer stack that starts with technical SEO and ends with commerce protocols. I recommend every SEO lead read it before their next quarterly planning.<\/p>\n The core idea: AI agents now browse, compare, and even purchase on behalf of users. If your site isn\u2019t structured for machines, you lose the sale before a human ever clicks. The stack covers crawlability (foundation), content clarity, off-site entity signals, task-completion readiness, and protocol support like WebMCP and Universal Commerce Protocol. Concrete dates back up the urgency\u2014ChatGPT\u2019s in-chat purchasing went live in February 2026, and the WebMCP draft landed in Chrome Canary last month.<\/p>\n I especially value the practical measurements. The article shows how to use Semrush\u2019s Site Intelligence to surface issues that agents hit, like broken links or JavaScript-rendered content. That\u2019s a gap few teams check today. The five-layer framework gives you a simple priority order: fix the foundation first, then layer on agent-friendly markup, then add protocol hooks. No fluff, no hype. If you focus on only one GEO tactic this quarter, make this the blueprint.<\/p>\n \ud83d\udd17 Semrush Blog<\/a><\/p>\n Zero-click searches hit 69.5% in the UK and just 62.1% in Germany \u2014 that\u2019s a 20%+ click-rate gap between these two markets.<\/p>\n Based on SparkToro and Similarweb\u2019s 2026 panel data across six countries, I think the EU\u2019s anti-self-preferencing rules are the most likely driver keeping German searchers clicking more. France stands out with the highest session-end rate (42.3%), meaning French users find answers faster and leave. For global SEO strategy, these country-level differences are critical. Germany still sends 287 clicks per 1,000 searches to the open web; the UK sends only 232. I recommend prioritizing organic investment in Germany and Italy, where clicks are higher. In the UK and US, expect more traffic to stay within Google\u2019s ecosystem. Bottom line: zero-click searches are the new normal everywhere, but the variance is large enough to shift budget allocation.<\/p>\n \ud83d\udd17 SparkToro Blog<\/a><\/p>\n This article is essential reading for anyone tracking both search rankings and AI citations, because it reveals why your dashboard numbers may be misleading you. Duane Forrester dismantles the assumption that feeding the same query to Google and an LLM produces comparable data. The core insight: a search index matches a string, while an LLM interprets one. Those are fundamentally different operations, and they reward different input shapes.<\/p>\n Key data points drive the point home. ChatGPT prompts run 17x longer than Google queries by character count. Models break those long prompts into multiple short retrieval queries\u2014clickstream analysis shows a 23-word typed prompt becomes a roughly 4-word search. Another study found more than two searches per prompt at about 5 words each. You are tracking the model\u2019s paraphrase of your query, not your original string.<\/p>\n I see a measurement trap here. One client writes tracked queries as tight noun phrases; another writes full conversational questions. Both have identical real-world visibility. But the first dashboard reads weak on both sides, the second strong. The report quietly converts a stylistic habit into a competitive gap.<\/p>\n I recommend auditing your tracked query set immediately. Strip out conversational long prompts that inflate citation numbers. Standardize to the actual text users type into a search box. Only then can you compare rank and AI citation honestly. This article gives you the mechanism to fix a broken report.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n The UK’s CMA now mandates that Google must give<\/strong> advance notice before significant ranking updates. This changes how every SEO plans for volatility.<\/p>\n I recommend reading this. Two new conduct requirements reshape UK search operations. First, Google must rank organic results objectively, including AI Overviews. Second, the data portability API becomes a legal obligation. Google has six months to comply. The CMA will monitor through regular reporting.<\/p>\n The fair ranking requirement targets a long-standing pain point. Businesses told the CMA that Google’s ranking changes arrive without warning. No effective complaint process exists. Now Google must provide transparent criteria and a formal route for concerns.<\/p>\n I think this marks a turning point. UK SEOs gain predictability. We can plan around ranking shifts rather than react to them. The requirement covers AI Overviews too. That brings generative search under the same fairness rules.<\/p>\n For practitioners outside the UK, watch closely. This sets a precedent for other markets. The US and EU are already increasing regulatory scrutiny.<\/p>\n Note one limitation: Google does not disclose the ranking algorithm. The requirement covers criteria, notice, and complaints. Implementation determines real value.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n This article from Search Engine Journal discusses how Google explains that URLs blocked by robots.txt can still be indexed.<\/p>\n John Mueller confirms the Search Console warning is often harmless. Add-to-cart URLs like \ud83d\udd17 Search Engine Journal<\/a><\/p>\n Google says markdown is a bad idea for AI SEO. This Search Engine Journal article summarizes a Search Off the Record podcast where John Mueller and Martin Splitt directly challenge the markdown trend.<\/p>\n I think this is essential reading for anyone optimizing for AI search. The core argument is simple: converting HTML to text is trivial. Google has done it for decades. Markdown adds no value for crawling or indexing.<\/p>\n What markdown removes matters. It strips navigation, headers, and internal links. These elements help search engines discover content and understand site structure. Without them, you lose context. Splitt explicitly says markdown makes discovery harder.<\/p>\n There’s also a trust problem. Search engines will never trust markdown as canonical content. They can easily extract original HTML. Relying on markdown is risky.<\/p>\n My take: SEOs pushing markdown miss fundamentals. Google wants complete HTML pages with structure. Stay with standard web pages. Don’t strip out the parts that help search engines understand your site.<\/p>\n \ud83d\udd17 Search Engine Journal<\/a><\/p>\n I think this is the most actionable YouTube keyword framework I’ve seen in months.<\/p>\n Phil Nottingham directly solves a pain point: how to quantify YouTube keyword opportunities without reliable search volume data. He introduces three original metrics \u2014 median monthly view velocity, median subscribers, and median age \u2014 as direct analogues to traditional SEO signals. The core insight: high opportunity means high MMVV, low median subscribers, and long median content age. I recommend using this framework to filter keyword lists before producing any video. It cuts through the guesswork and gives you a data-backed signal of where your channel can actually compete.<\/p>\n
\nImproving health intelligence in ChatGPT<\/h3>\n
\nNew OpenAI Academy courses for the next era of work<\/h3>\n
\n\ud83e\udd16 GEO\u00b7SEO Highlights<\/h2>\n
I made the same AI compete against itself in SEO tasks<\/h3>\n
\nGoogle’s Updated Guidance Now Says It’s “Fine” To Use LLMs.txt For AI SEO<\/h3>\n
\nWhy category entry points belong in every AI search strategy<\/h3>\n
\nWhat Your CMO Needs to Know About AI Search w\/ Tom Critchlow<\/h3>\n
\n
\nWhy Calling Yourself The Best Could Be Helping Your Competitors Win In AI Search<\/h3>\n
\nAgentic Commerce For Small Merchants: Which Protocol Spec Actually Matters For Your Website<\/h3>\n
\n
\nWhy Publishing More Content Is Making Your SEO Worse<\/h3>\n
\n11 Ways to Automate SEO with Agent A<\/h3>\n
\nAgentic Marketing: What’s the Big Deal and How to Get Started<\/h3>\n
\n10 SEO Trends I’ve Seen Firsthand in 2026 (With Data)<\/h3>\n
\nThe Integrated Search Brief That Aligns SEO, PPC & Content In The AI Search Era<\/h3>\n
\nWe Analyzed 137K Sites: 97% of llms.txt Files Never Get Read<\/h3>\n
\n
\nHow to optimize for the agentic web: a guide for marketers<\/h3>\n
\nZero-Click Searches: Highest in the UK, Lowest in Germany, and France has the Most Efficient Searchers<\/h3>\n
\nRank And AI Citation Aren\u2019t The Same Number<\/h3>\n
\nGoogle Must Give Notice Before Significant Ranking Changes<\/h3>\n
\nGoogle Explains Why URLs Blocked By Robots.txt Can Still Be Indexed<\/h3>\n
?add-to-cart=<\/code> don’t need indexing. Blocking them with robots.txt is fine. Google rarely shows those URLs in search results. Adding a noindex tag isn’t a practical fix because the parameterized page is the same as the regular product page. Instead, I recommend crawling your site with Screaming Frog. Find internal links pointing to those URLs. Add rel=\"nofollow\"<\/code> to those links. That gives Google a strong hint to stop crawling them. Mueller’s core takeaway: not every Search Console alert requires action. In this case, the internal links are the real issue. The robots.txt block already works well. Focus on link hygiene rather than changing indexing directives.<\/p>\n
\nGoogle Says Markdown For AI SEO Strips Away The Parts That Matter<\/h3>\n
\nQuantifying YouTube Keyword Opportunities \u2014 Whiteboard Friday<\/h3>\n