Krishna Madhavan - Beyond Blue Links: Freshness, Control, and Visibility

Transcript

1. Beyond Blue Links: Freshness, Control, and Visibility Krishna Madhavan Principal Group Manager, Microsoft Bing
2. Cats still rule the Internet!
3. At Microsoft, I drive the collection and understanding of the web to drive experiences across Bing, Copilot, and beyond. more… Search & Ads News Images/Videos Copilot Gen Search more… { {
4. This is what the future of SEO/AEO is about. Visibility belongs to content that AI can trust, understand, and ground.
5. AEO and Grounding - All about visibility For AEO, grounding is the connective tissue between content, and the trust signals AI engines require for visibility
6. Grounding Controls Freshness Demystify & connect 3 key visibility concepts
7. Grounding Anchors an LLM’s output to verifiable, retrievable, trusted sources before generation
8. Anatomy of grounding
9. Simplified view: The AI answer chain (1/2) User query User submits a query or a task Retrieval (multi-source) Relevant content is fetched from available sources (usually the index). Safety starts even at this stage. Ranking Retrieved content scored & ordered by relevance, quality, authority, freshness Evidence synthesis Top-ranked items are merged, de-duplicated, summarized, and structured into clean evidence blocks. Safety filtering Evidence is checked against safety, compliance, and policy constraints (PII, harmful content, etc.)
10. Factual alignment The model ensures that generation stays anchored to the selected evidence (reducing hallucinations). Context integration The appropriate user, application, and situational context is integrated. Answer Generation The grounded answer is generated using the synthesized, filtered, aligned evidence. This step receives a lot of focus in AEO/SEO. Citations Source attribution is produced, mapping factual statements back to their originating source. Simplified view: The AI answer chain (2/2)
11. New visibility funnel In AEO, visibility starts long before generation with query understanding. User query is interpreted, normalized, and fanned out into multiple retrieval paths.
12. User Query (only illustrative purposes) “What are the cutest new cat toys this year?”
13. Query is vague, informal, and incomplete The system cannot use this query as is 01 Examine the query
14. AI interprets the query (example) topic: cat toys attribute: new (freshness-sensitive) desired outcome: recommendation tone: positive, fun/consumer intent Visibility starts here: If your content doesn’t match the interpreted intent, it is likely never considered. 02 Query understanding
15. AI rewrites the query into a clean, explicit version it can operate on: “latest cat toys for 2025” “new cat toy releases for this year” “top trending cat toys” First selection bar: Only content aligned with the normalized intent is usually retrieved. 03 Query normalization
16. The system generates multiple retrieval paths. Fanout = many parallel versions of your intent Semantic, keyword, schema-based, attribute-specific, entity, domain, and freshness queries All competing to find your content. 04 Query expansion (1/2)
17. 04 Query expansion (2/2) Fanout Type Example Variants Semantic trending cat toys 2025; popular new cat toys Keyword new cat toys this year; latest cat toy releases Freshness recently released cat toys; 2025 cat toy launches Schema/Structured Product: cat toy (releaseDate=2025) Attribute interactive cat toys; enrichment toys Category/Entity new kitten toys; toys for playful cats “What are the cutest cat toys this year?”
18. Simplified view: The AI answer chain (1/2) User query User submits a query or a task Retrieval (multi-source) Relevant content is fetched from available sources (usually the index) Ranking Retrieved content scored & ordered by relevance, quality, authority, freshness Evidence synthesis Top-ranked items are merged, de-duplicated, summarized, and structured into clean evidence blocks. Safety filtering Evidence is checked against safety, compliance, and policy constraints (PII, harmful content, etc.)
19. Simplified view: The AI answer chain (1/2) User query User submits a query or a task Retrieval (multi-source) Relevant content is fetched from available sources (usually the index) 01 Examine the query 02 Query understanding – Visibility starts here! 03 Query normalization 04 Query expansion
20. Making content modular, multi-intent, structured, and easily crawlable helps with grounding – therefore, visibility AUTH EN TIC IT Y F RE SH NE SS S EM AN TIC R IC HN ES S
21. Freshness influences retrievability, authenticity influences trust scoring, and semantic richness influences match quality. Signal Stage Impact on Grounding Freshness Retrieval (2) Surfaces up-to-date content for recency-sensitive queries Ranking (3) Boosts newly updated, relevant pages Evidence Synthesis (4) Prefers latest facts particularly when conflicts exist Authenticity Ranking (3) Rewards credible, expert, original sources Safety Filtering (5) Removes untrustworthy or spammy pages Factual Alignment (6) Increases confidence in evidence used for grounding Semantic Richness Query Understanding (1) Better match to expanded intents & fanout paths Retrieval (2) Stronger semantic matches through embeddings Evidence Synthesis (4) Provides clearer, more extractable facts
22. Content freshness drives AEO visibility If your information is stale, grounding will probably not select it, no matter how well-written it is.
23. Freshness is about super fast detection of New content Content with meaningful change Deleted content
24. Inefficiency in crawling affects content freshness In most AI and search crawling, billions of URLs with no content change are crawled everyday. No content change Still crawling Wasted crawls
25. Balancing crawl efficiency vs. content freshness is a challenge for crawlers…. especially, for long-tail and low-signal sites Crawl efficiency Content freshness
26. Crawl Efficiency → Meaningful crawls at scale Focus on new, updated, deleted pages Meaningful update: changes to primary text, images, schema, images, videos, outlinks
27. IndexNow – AEO Visibility Smarter way to reduce crawl load, increase efficiency, and significantly enhance content freshness • Open Protocol – Adopted by a rapidly growing community of 10s of millions of websites. • Timely and clear freshness signal – Enable instant notification of new, updated, or deleted content • No wait – Take control of crawling. No need to wait for crawler to visit. Reduce number of crawls!
28. IndexNow – Just notify us of content change! Take control of crawling! Partners instantly notified (Yandex, Naver, Yep, Bing, Seznam.cz, Amazon) No wait! Fast crawls! Ensure freshness! New, updated, deleted content Indexnow API
29. IndexNow Impact Significant impact through content freshness, crawl efficiency 50 % Clicked, newly indexed URLs in Bing SERP results originate from IndexNow We do NOT favor IndexNow URLs in SERP. However, we do focus on content freshness over legacy content!
30. Our long-term goal: To make IndexNow more efficient, easier to adopt, and environmentally friendlier! Credit Ahrefs: https://ahrefs.com/index-now
31. IndexNow: Path Forward Please act now: • Learn more: www.indexnow.org • Learn about IndexNow partners: www.bing.com/indexnow Our commitment: We strive to make crawling efficient – especially for IndexNow-enabled sites.
32. Authenticity & Semantic richness Freshness pulls content into the results; authenticity and semantic richness push it into the answer.
33. Top 5 Content Tips for authentic, semantic rich content (1/2) 01 Build unique “content depth intensity” Focus on areas that reflect your unique expertise, depth, and niche problem space. 02 Structure is part of storytelling in “chunks” Avoid <div> soup, use <h1> or <h2> to position key points, tables, FAQ blocks, schema markup, and think “chunks”. 03 Showing your work builds trust Support your claims with examples, data, and sources. Content grounded in evidence builds trust and well- anchored claims.
34. Top 5 Content Tips for authentic, semantic rich content (2/2) 04 Context rules Go beyond facts. Explain why it matters, when it matters, and what to watch for. Truly understand your users. 05 Freshness is not a gimmick – it is governance Treat content refresh, consistent update cadence as part of knowledge lifecycle management.
35. Controls Can we control what the AI is allowed to see and use?
36. AEO Content Controls AI systems make 2 distinct decisions: 1. Access: Can content be retrieved? robots.txt, noindex, enterprise permissions 2. Use: Can content be used? data-nosnippet, no-snippet, snippet limits, schema-scoped boundaries
37. Blunt vs. Precision Controls 1. Blunt control: robots.txt, noindex, blocking crawlers • Removes your entire page from eligibility • No visibility, ranking, and grounding opportunities 2. Precision control: data-nosnippet • Handy HTML attribute for fine-grain control
38. AEO Precision Content Controls data-nonippet HTML attribute to stay in the index, ranking, but excluded from snippets and AI summaries. Control what is shown in previews without affecting page’s visibility or ranking potential. See our blog: data-nosnippet
39. <h2>Overview</h2> <p>This section may appear in Bing Search and Copilot answers.</p> <div data-nosnippet> <h3>Subscriber Content</h3> <p>This section does not appear in Bing Search and Copilot answers.</p> </div> data-nosnippet - HTML attribute usage example <h2>O v er vi ew </h 2> <p>Th i s s ec ti on m ay app ear i n Bi ng Sea rc h a nd Copi l ot ans w er s. </p > <di v dat a-nos ni ppet > <h3>S ubs cri be r Cont ent </h3 > <p>Th i s s ec ti on does not ap pear i n Bi ng Se arc h and Copi l ot an sw er s. </ p> < /d i v >
40. Use Case What It Prevents Why It Matters for AEO Hide answer to trigger clicks Specific portion surfacing Hides specific part to trigger clicks and experiment Protect context-dependent content Nuanced instructions, expert-only guidance Prevents misinterpretation by LLMs Safe A/B testing Experimental copy, variant messaging Test content without exposing it in snippets/AI Shield proprietary/premium info Paid or gated material Keeps ranking while blocking evidence extraction Control factual grounding Specs, KPIs, internal metrics Decide which facts AI can turn into evidence Block tables/lists/specs High-value extractable structures Prevent automatic grounding from structured blocks Remove noisy/ambiguous blocks Fluff, marketing hype, unclear text Improves semantic clarity & reduces grounding errors Precision control with data-nosnippet
41. Directive What It Does Impact noindex Removes page from search index No visibility; cannot be used in grounding. data-nosnippet Blocks specific HTML sections from snippets/grounding Preserves ranking; hides only selected fragments Visibility Matters: Choose controls that protect it
42. www.bing.com/webmasters Bing Webmaster Tools (Free) ✓ Improve Search Visibility ✓ Optimize for Performance ✓ Fix Issues & Enhance Indexing Get your site's performance data, take advantage of free SEO tools, and analyze insights to improve your page rankings in search and Copilot results.
43. Microsoft Clarity (Free) clarity.microsoft.com ✓ Heatmaps for all pages ✓ Session recordings ✓ Insights Gain deeper insights into user behavior, track engagement, and identify opportunities to optimize your site for better conversions and a seamless experience.
44. Freshness, authentic, semantically rich content – combined with precision controls Help AI do what cats do best: stay curious, balanced, and graceful!
45. Beautiful images – MAI- Image-1 with Bing Image Creator Thank you! Dr. Krishna Madhavan Principal Product Manager, Microsoft Bing LinkedIn: https://www.linkedin.com/in/kmadhavan7/