Google Closes the AI Overview Index Gap

What indexing lag did Google just remove?

Google appears to have closed the delay between deindexing a page in Search and removing it from AI Overviews and AI Mode. Glenn Gabe noted, per Search Engine Roundtable in late May, that a site hit by a manual action and deindexed no longer showed in AI surfaces at all, suggesting Google aligned them.

Why does index alignment matter for AI visibility?

It means AI Overviews, AI Mode, and classic Search now draw from one aligned index. The same crawlability and quality work feeds all three surfaces at once. You no longer manage AI visibility as a separate track. If a page ranks and stays indexed, it stays eligible across every Google surface together.

How does this change AEO strategy?

It simplifies it. You stop chasing AI-specific shortcuts and focus on the fundamentals that already work: crawlable pages, clear answers, and real authority. Since the surfaces share an index, strong technical SEO and trustworthy content now serve answer-engine optimization directly. The takeaway is that AEO and SEO are converging, not splitting apart.

What did Gary Illyes warn about buying mentions?

At Search Central Live Shanghai on May 15, Gary Illyes cautioned against buying or manipulating brand mentions to game AI systems, per Digital Phablet. He framed paid mentions as similar to paid link schemes. The message is that artificial mention-building is a risk, not a reliable path to AI citations or visibility.

Should you stop chasing AI-mention shortcuts?

Yes. Paying for mentions to inflate AI visibility carries the same risk profile as buying links, and Google has signaled it. The durable approach is earning genuine mentions through useful content, real coverage, and a credible brand. Shortcuts that manufacture mentions can backfire, and they do not build the lasting authority AI engines reward.

When will deindexed pages drop from AI answers?

Effectively right away now, rather than lingering for days. Before this change, observers saw deindexed sites still appear in AI Overviews or AI Mode for a short window. With the surfaces aligned, removal from Search now seems to remove a page from AI answers in step, closing that gap.

Which signals still drive AI citations?

The same ones that drive search visibility: clear, well-structured content, demonstrated expertise, and crawlable pages. Google's guidance points to normal SEO practices working for AI surfaces too. Direct answers, question-based headings, and verifiable facts help, because the model needs a clean, trustworthy passage to lift. There is no separate secret AI ranking factor.

Will llms.txt help in Google?

No. Google has said it does not use llms.txt files for Search, and that position extends to its AI surfaces. Adding one will not improve your Google AI visibility. The file may serve some third-party agentic tools, but for Google specifically, it does nothing. Spend that effort on content and crawlability instead.

Can manipulated mentions get you penalized?

They carry real risk. Google equates buying mentions with paid link schemes, which can draw manual actions. A manual action can deindex you, and with the surfaces now aligned, that removes you from AI Overviews and AI Mode too. So manipulation does not just fail to help, it can pull you out of every Google surface.

Where should a SaaS team focus instead?

On crawlable, well-structured pages that answer real buyer questions and earn genuine authority. Keep your technical SEO clean, publish content with clear answers and sources, and build real coverage. Because Search and AI surfaces share an index, that single investment now pays off everywhere Google shows your brand. Fundamentals are the strategy.

Aligning your AI and search strategy? Pair this with my guide to AEO, GEO, and SEO, the Reddit AI citation index piece, and the breakdown of Google's llms.txt mixed signal. Let's chat.

For the newer Highlighted Answers feature that sits on top of AI Mode, my newer note on Google AI Mode Highlighted Answers for Webflow pages covers the schema and answer-block pattern that wins one.