SEO News: February 2026

Yahoo Scout: The New AI “Answer Engine”

Yahoo has officially launched beta testing for Yahoo Scout – an AI-powered search engine. Unlike a traditional list of links, Scout synthesizes information to provide direct, structured answers backed by transparent sources.

• Leveraging 30 Years of History: Scout draws from Yahoo's unique datasets, which include 500 million user profiles and a knowledge graph with over a billion entities.
• Portfolio-Wide Integration: The new Yahoo Scout Intelligence platform is being embedded across all services – from Mail and News to Finance and Sports (e.g., automatic email summaries or in-depth game result analyses).
• Yahoo Shopping – Your Shopping and Financial Assistant: It can compare products and expert reviews in seconds, replacing hours of manual searching.
• Yahoo Finance: Offers real-time market analyses that update every 10 minutes.
• Strategic Partnerships: The primary model for Scout is Anthropic’s Claude (chosen for its speed and safety).
• Open-web data is provided by Microsoft Bing via their grounding API.

The service is currently available in beta for users in the U.S. (desktop and mobile apps). In the coming months, Yahoo plans to introduce deeper personalization and launch new generative search ad formats for advertisers.

Google AI Overviews: The End of Clicks? Now "Switching" You Into AI Mode

Google is introducing a significant change in how users interact with AI Overviews. The goal is to keep users within the Google ecosystem instead of sending them to external websites.

• Direct Jump to AI Mode: If a user asks a follow-up question within an AI Overview, Google now transitions them directly into the full-featured AI Mode. According to Google, this transition is more natural and seamless.
• Loss of Sources and Clicks: In this new interface (currently on mobile devices), the original search results are overlaid by the AI interface. A critical issue for content creators is that the original source cards and links are missing in this view, which will likely lead to a further decline in organic traffic.
• Gemini 3 as the Standard: Google has confirmed that the Gemini 3 model is now the global default for all AI Overviews. This translates to higher quality and more complex answers delivered directly on the Google search page.

Google is increasingly aggressive in its push to transform from a search engine of links into an “answer engine.” For website owners, this conversational model means that securing clicks from the top of search results will be even more challenging than before.

Bing Webmaster Tools is testing a new “AI Performance” report

Microsoft has begun fulfilling its promise through a very limited beta test, displaying specific AI search performance data within Bing Webmaster Tools.

• Citation Data: The new report focuses on how often your website is cited in responses from Microsoft Copilot and its partners.
• Report Metrics: It displays the number of citations and the number of cited pages over time (on a daily basis).
• Query Interpretation (Grounding Queries): The report does not show the users' literal queries; instead, it displays how Bing interpreted them, including intent classification – whether the query is informational, navigational, etc.
• The Catch: The report currently does not display the number of clicks or the click-through rate (CTR). This means you can see that the AI "used" your site as a source, but you don't know if it actually drove a real visitor to your website.

This represents the first step toward transparency in how websites rank within Bing’s AI responses. However, the absence of click data suggests a reluctance by search engines to admit how much AI search is cannibalizing website traffic.

AI citations are extremely inconsistent

A new SparkToro study has revealed that relying on AI recommendations is a risky bet for marketers. This is because AI tools respond differently to identical queries every single time.

• Less than 1% Match: The chances of ChatGPT or Google AI generating the exact same list of brands or products twice in a row for the identical query are lower than 1%.
• Random Order and Quantity: AI varies not only which brands it recommends but also their ranking order and the total number of items in the list (sometimes providing 3, other times 10).
• User Variability: People phrase their queries in vastly different ways. Even when they share the same intent (e.g., “find the best headphones”), the semantic similarity of their prompts is minimal, further increasing the chaos in search results.
• The End of “Rank Tracking” in AI: Traditional position tracking (1st place, 2nd place, ...) makes no sense in the AI era. If the result changes with every refresh, there is no such thing as a stable “ranking.”

Instead of tracking positions, it is more effective to measure “Visibility Percentage.” This means calculating how often your brand appears at all within a large sample (e.g., 100 queries). In the era of AI search, this is the only metric with real informative value.

Googlebot File Size Limits

Google has clarified the exact amount of data its bots actually crawl. This is not a change in search engine behavior, but an important refinement of figures that were previously reported inconsistently.

• General Infrastructure Limit (15 MB): Most of Google’s crawlers and fetchers (for Shopping, News, Gemini, etc.) have a default limit of 15 MB per file.
• Googlebot and Google Search (2 MB): For standard web search indexing of HTML and supported text-based files, the limit is set at 2 MB.
• PDF Files (64 MB): For documents in PDF format, the limit is significantly higher, at 64 MB.
• Independent Resource Loading: Crucially, these limits apply to the individual file itself (e.g., the HTML source code). External resources such as CSS and JavaScript are downloaded and metered independently.

If your HTML page (source code excluding images) exceeds the 2 MB limit for Search—or the 15 MB general infrastructure limit—Googlebot will stop downloading further data. Any content located in the code beyond this threshold will not be indexed.

In the era of AI and complex scripts, it is easy to "bloat" HTML code. This update serves as a reminder that for proper indexing—especially for content-rich pages—it is critical to keep the code clean and move non-essential elements (such as massive inline scripts) into external files.

Google Ads in 2026: What to Expect

In 2026, Google is rewriting the rules of digital marketing. The objective is to create a shopping experience that is seamless, personalized, and driven by intelligent agents.

YouTube: From Inspiration to Purchase

• AI Matching: Google leverages AI to deeply analyze content and audiences, instantly connecting brands with the most relevant creator communities.
• Business Impact: The goal is to shorten the user journey from viewing a video (“just looking”) to completing a purchase (“already have it”).

Ads in AI Mode

• New Formats in AI Mode: Google is testing sponsored ads directly within the conversational AI interface. These ads appear alongside organic recommendations, helping users compare brands and stores in real time.
• Direct Offers: At the exact moment a customer is ready to buy, businesses can provide a personalized offer—such as a discount, loyalty bonus, or bundle—without changing the price for other users.

Agentic Commerce and UCP/AP2 Protocols

• Shopping Agents in Action: Agentic commerce is becoming a reality. Google is aiming to remove the "drudgery" of shopping by letting AI agents act on behalf of the user.
• Universal Commerce Protocol (UCP): This standard enables a secure connection between businesses and AI agents, incorporating digital identity verification and automated payment processing.
• Direct AI Checkout: U.S. users can now purchase items from merchants like Etsy and Wayfair directly within AI Mode or the Gemini app without leaving Google’s interface. Support for Walmart, Target, and Shopify is currently rolling out.

Creativity and Performance with Gemini 3

• Asset Studio: Tools like Nano Banana 2 and Veo 3.1 allow advertisers to create studio-quality images and videos in minutes. In 2025, the volume of AI-generated assets increased threefold, fundamentally shifting creative workflows.
• AI Max: A new campaign suite (the successor to Performance Max for Search) that helps advertisers reach billions of net-new search queries that were previously uncovered. It leverages keywordless matching and intent-based signals to capture incremental demand in real-time.

Google emphasizes that the entire "AI-first" search era is built upon rigorous data protection standards. Even as agents act on behalf of users, security remains a top priority.

OpenAI Launches Ad Testing in ChatGPT

OpenAI officially launched (February 9, 2026) testing of its advertising system directly within the ChatGPT interface. The objective is to secure funding for free access to advanced models for hundreds of millions of users.

• Target Audience: The test is currently running in the U.S. for adult users on the Free and Go (budget-friendly) tiers. Subscribers of Plus, Pro, Business, and Enterprise remain completely ad-free.
• Placement and Format: Ads are visually separated from responses and clearly labeled as “Sponsored.” They appear as supplementary cards relevant to the conversation topic (e.g., a search for recipes might trigger a grocery delivery offer).
• Response Independence: OpenAI guarantees that advertisers cannot influence ChatGPT’s generated responses. Answers remain impartial, prioritizing user value over commercial interests.

Privacy and Control

• Conversation Privacy: Advertisers do not have access to chat histories, memories, or personal data. They receive only anonymized, aggregated data regarding impressions and clicks.
• Targeting: Ads are served based on the current conversation topic and previous interactions, but they are strictly excluded from sensitive topics (health, politics) and for users under the age of 18.
• User Choice: Users can dismiss specific ads, delete their advertising data, or opt out of personalization entirely—in exchange for lower limits on free messages.

For businesses, this opens a brand-new channel to reach users at the exact moment they are actively considering a purchase decision. OpenAI plans to gradually expand the system with additional formats and interactive ways to connect brands with consumers.

Markdown Files: Cloudflare offers an automated solution to serve AI-optimized content formats.

Cloudflare has launched a new feature called "Markdown for Agents," enabling websites to automatically serve their content to AI bots and agents in Markdown format instead of traditional HTML.

• Token Efficiency (Up to 80% Savings): HTML code is "clutter" for AI. While a standard HTML article can consume over 16,000 tokens, its Markdown version takes only about 3,000. This dramatically reduces AI operational costs and accelerates response times.
• Structure over Visuals: AI agents are indifferent to sidebars, banners, or scripts. Markdown provides a clean semantic structure (headings, lists, tables) that agents can process instantly without the risk of conversion errors.
• On-the-fly Automatic Conversion: When enabled, Cloudflare detects requests from AI agents (via the Accept: text/markdown header) and serves optimized Markdown directly from its edge network in real-time.

• Content Signals: Cloudflare adds metadata to headers indicating whether the author permits the use of their content for AI training, search indexing, or as input for autonomous agents (e.g., ai-train=yes, search=yes, etc.).
• X-Markdown-Tokens: A specialized header that informs the agent exactly how many tokens the document contains. This helps AI better manage its memory and plan its context window usage.

The feature is already utilized by popular developer agents such as Claude Code and OpenCode, and is fully supported by OAI-Searchbot (the search crawler for ChatGPT).

ChatGPT vs. Google: 12% of Search Volume, but 190x Less Traffic for Websites.

The relationship between traditional search engines and AI assistants is undergoing a dynamic transformation. Recent data from Ahrefs shows that while ChatGPT successfully competes with Google in search volume, its role in the website traffic ecosystem is dramatically different. While Google serves as a "gateway," ChatGPT acts more like a "destination."

• Query Volume: ChatGPT processes approximately 2.5 billion prompts daily. This represents 18% of Google's total daily search volume.
• Search vs. Chat: Approximately 65% of interactions in ChatGPT are classified as queries that users previously entered into Google.
• Traffic Gap: Google sends 190 times more traffic to web pages than ChatGPT.
• Click-Through Rate (CTR): ChatGPT has a 96% lower CTR than Google. Users receive answers directly within the interface and no longer click through to source websites.

As of January 2026, Google continues to dominate, but ChatGPT has already surpassed traditional search engines like Bing in query volume (Bing is estimated at 1.2 billion queries daily).

Google will improve the visibility of links in AI Overviews.

Google continues to refine its AI search results. Robby Stein, Vice President of Product at Google Search, announced on X a redesign of how source links are displayed within the AI Overviews and AI Mode interfaces. The goal is to increase transparency and make it easier for users to click through to the original content.

• Hover Pop-ups (Preview on Hover): On desktop, hovering over a link within an AI response now triggers a pop-up window. This includes the website name, favicon, and a brief description, allowing users to immediately see the source of the information.
• Prominent Icons: Google is introducing more visible and descriptive link icons directly within the response text, on both desktop and mobile versions.
• Better User Engagement: According to Google's internal testing, this new user interface (UI) is more "engaging" for users and leads to more frequent click-throughs to websites.

With these iterations, Google is attempting to find a balance between providing rapid answers and maintaining traffic for content creators. If it is confirmed that a more "engaging" design truly leads to higher click-through rates, it could alleviate concerns regarding so-called "zero-click" search.

Google Search Console: AI Performance Filtering

Google has officially concluded testing and globally deployed a new AI-powered tool directly within Google Search Console (GSC). This feature allows users to manage performance reports using prompts instead of manually clicking through filters.

Instead of complex parameter settings, you simply describe the analysis you want to see. The AI instantly transforms your request into the corresponding filters and report settings.

• Metric Selection: Based on your query, it selects relevant data (Clicks, Impressions, CTR, or Average Position).
• Filter Application: It can filter data by query, specific page, country, device, or search type.
• Comparison Settings: It easily configures complex time comparisons, such as custom date ranges, without the need for manual input.

Currently, this feature is available only for Search results (not for Discover or Google News). Although it is an AI that may not always be 100% accurate, it significantly speeds up data analysis and helps uncover correlations that might be overlooked during manual filtering.

Try asking, for example, for a "comparison of mobile clicks over the last 3 months versus the previous year" and see how quickly the AI prepares the report.

Markdown Files in SEO: Google and Bing Warn of Complications

A debate has recently flared up within the community regarding the use of Markdown files (.md) for SEO purposes. Although some believe that a simpler format may make work easier for AI bots, representatives from the largest search engines, John Mueller (Google) and Fabrice Canel (Bing), remain skeptical of this trend.

• Crawl Budget Overload: Using separate Markdown versions of pages can increase crawl budget demands without any real benefit.
• Error Rates and Maintenance: Fabrice Canel (Bing) pointed out the risk that the .md version of a page may be "half-broken" and left unmaintained, as the focus is solely on bots rather than humans.
• User-Centric Ranking: Both Bing and Google emphasize that their systems evaluate content based on what a real user sees (rendered HTML), rather than the raw text in Markdown.
• Risk of Spam and Manipulation: If SEO specialists begin using .md files on a massive scale to manipulate results, search engines will simply start filtering or ignoring them.

Analyst Glenn Gabe compares Markdown SEO efforts to the earlier AMP format. With one crucial difference: while AMP offered clear speed benefits and special labeling in search results, there is no confirmed reward from Google or Bing for using Markdown files.

Why ChatGPT Can't Find You: The Language Barrier You Didn't Know About

A new analysis of over 10 million ChatGPT queries has revealed a troubling fact for local businesses: even when you enter a query in Czech, Polish, or German, the AI searches for a large portion of its supporting information in English. As a result, local market leaders often vanish in the shadow of global brands.

How ChatGPT "Thinks" Behind the Scenes

When you submit a query, ChatGPT splits it for internal purposes into several sub-queries (so-called query fan-outs). Analysis has shown that:

• 43% of all research occurs on the English-language web, even for queries submitted in other languages.
• In 78% of sessions in foreign languages, ChatGPT performs at least one sub-query in English.
• Language Bias: For Spanish, 66% of research is conducted in English, while for Turkish, it is as high as 94%.

Impacts on Local Business

This mechanism creates a systemic advantage for global players. In practice, it looks like this:

• Poland: When querying the best auction portals (in Polish from a Polish IP address), ChatGPT often de-prioritizes the local giant Allegro and prefers eBay.
• Germany: A query about German software companies often generates a list of international brands because the sources ChatGPT cites are in English.
• Spain: For cosmetics, ChatGPT recommends global brands because it automatically adds the word "global" to the query during its research.

Why is this happening?

The reasons are not intentionally harmful, but technical:

1. Authority Signals: Global (English) sites have more backlinks and citations, which models perceive as higher trustworthiness.
2. Risk Minimization: Half of the internet's content is in English. For AI, it is safer to search where there is the highest probability of finding a "quality" source.

"ChatGPT's language bias is not malicious intent, but a consequence of training data and the dominance of English on the web. As a result, local businesses are losing out even in their home markets," states Tomek Rudzki.

What to do about it? Local SEO Strategy

If you want to be visible in AI search, a mere translation of your website is not enough. Rudzki recommends:

• Citation analysis: Find out which English-language sources (Reddit, Wikipedia, industry magazines) ChatGPT cites in your segment.
• Strategic English content: Create English versions of key pages or aim for mentions in English-language "listicles" from which AI draws its information.
• Competitor monitoring: Track which English-language sources your competitors appear in while you do not.

Until strong multilingual AI models trained primarily on local data emerge, companies must adapt. Being a local market leader in 2026 means being visible even to the "English eyes" of artificial intelligence.

Sources

• https://www.yahooinc.com/press/introducing-yahoo-scout-a-new-ai-answer-engine
• https://searchengineland.com/google-ai-overviews-follow-up-questions-jump-you-directly-to-ai-mode-468016
• https://searchengineland.com/bing-webmaster-tools-testing-new-ai-performance-report-468039
• https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview
• https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/
• https://www.searchenginejournal.com/google-updates-googlebot-file-size-limit-docs/566485/
• https://blog.google/products/ads-commerce/digital-advertising-commerce-2026/
• https://openai.com/cs-CZ/index/testing-ads-in-chatgpt/
• https://blog.cloudflare.com/markdown-for-agents/
• https://ahrefs.com/blog/chatgpt-has-12-percent-of-googles-search-volume/
• https://www.searchenginejournal.com/google-says-links-will-be-more-visible-in-ai-overviews/567719/
• https://x.com/rmstein/status/2023865995908313531?s=20
• https://searchengineland.com/google-search-console-ai-powered-configuration-rolling-out-469349
• https://www.seroundtable.com/google-bing-markdown-files-40928.html