Google Warned Against Content Chunking at Milan — Here's What It Actually Means

I've been watching the SEO industry twist itself into knots trying to optimize for AI search — and honestly, some of it deserved what Google said at Search Central Live in Milan this week.

The event covered a lot of ground: site-wide signals, paywalls, AI blocking, how AI search handles clicks differently from organic. But the piece that lit up every SEO Slack channel I'm in was Google's stance on content chunking. Specifically: stop doing it artificially if you care about ranking in traditional search.

Let me break down what was actually said, why there's more nuance here than the hot takes suggest, and what you should actually do about your content strategy right now.

What "Content Chunking" Actually Is (And Why People Started Doing It)

If you've spent any time in SEO Twitter over the past year, you've seen the playbook spreading. It goes something like this: AI systems like ChatGPT, Perplexity, and Google's own AI Overviews extract passages from web pages. So the thinking was — if you break your content into tidy, self-contained bite-sized blocks, each answering one specific question, you'd make it easier for AI to pick up your content and cite you.

The result? Pages that look like this:

You've seen these pages. They feel hollow. You click, scan three bullet points, realize there's no actual explanation behind any of them, and close the tab. But the theory was that AI wouldn't notice — it would just lift the clean little blocks and cite your site.

Google noticed.

What Google Actually Said at Milan

The Search Central Live event in Milan this week covered several things, but the chunking advice was pretty direct: don't artificially restructure your content into bite-sized pieces just for LLMs. Write for people. Structure naturally. Google's systems are sophisticated enough to understand and extract relevant passages from well-written, human-readable content — you don't need to pre-chunk it.

There's an obvious irony here that people immediately jumped on: Google's own AI Overviews literally extract and stitch together chunks of content from multiple sources. So Google is doing the chunking for AI. They just don't want you to do it preemptively in a way that degrades the actual reading experience.

This is actually a consistent message from Google going back several algorithm cycles. Quality is determined by whether your content genuinely serves the person reading it, not whether it's been formatted to be machine-extractable. The difference is: good structured content helps both humans and AI. Content pre-fragmented to game AI extraction often helps neither.

The Site-Wide Signals Part Nobody's Talking About

The chunking discussion got the headlines, but honestly the site-wide signals discussion at Milan was more important for most SEOs.

Google reiterated something I think gets badly underestimated: they look at your site as a whole, not just individual pages. If a significant portion of your site is low-quality, thin, or — you guessed it — artificially formatted content with no real depth, it drags down the pages on your site that are excellent.

This is why content pruning keeps coming up. It's not just about removing bad pages. It's about ensuring that Google's assessment of your site as a whole is positive. One genuinely excellent in-depth article surrounded by fifty thin "chunked for AI" pages isn't a good site — it's one good page on a bad site.

The practical implication: if you've been publishing chunked, thin content across your blog because someone told you that's what AI search requires, you may have been quietly degrading your site's overall quality signal. Auditing and cleaning that up now is more urgent than it sounds.

Paywalls, AI Blocking, and the Traffic Question

Milan also covered a question that a lot of publishers have been wrestling with: should you block AI crawlers, and what happens to your traffic if you do?

Google's position here was pretty consistent with what they've said elsewhere: you can block AI crawlers. That's your right. But you can't selectively allow Googlebot for SEO purposes while blocking all AI access — that gets into manipulative territory, and Google is watching for inconsistency there.

The more interesting part was the discussion of "AI clicks" — how clicks from AI search results behave differently from traditional organic clicks. Users who click through from AI Overviews tend to be higher-intent; they've already gotten a summary, so if they click your site, they actually want something more. The Milan talk seemed to suggest that while raw click volume from AI surfaces may be lower, the quality of those clicks can be substantially higher.

What This Actually Changes for Your AI Search Strategy

Here's my honest take after absorbing everything from Milan: this isn't a pivot. It's a correction.

The people who went hard on "optimize for AI extraction" and sacrificed content depth in the process got burned twice — first on traditional rankings (because site quality dropped), second on AI citations (because AI systems actually prefer authoritative, detailed sources over fragmented ones).

The people who kept writing genuinely useful, in-depth content with clear structure? They're fine. They were always fine.

What the Milan event really confirms is that the gap between "good for traditional SEO" and "good for AI search" is much smaller than the content fragmentation crowd was claiming. The fundamentals haven't changed as dramatically as certain consultants have been charging clients to "pivot to GEO" would have you believe.

That said, there are real things you can do to improve your visibility in AI search results — and they're additive to good SEO, not a replacement for it.

The Practical Checklist: Write for Humans, Win in AI Search

Based on everything from Milan — and honestly from watching what's actually working across dozens of sites I track — here's what I'd focus on:

The Bigger Picture: Stop Chasing the Last Trend

There's a pattern in SEO that repeats every few years. Someone identifies a real signal or a real thing that AI/algorithms respond to. It spreads. People take it too far. Google corrects. Everyone scrambles.

Content chunking for AI is the latest version of this. The original observation wasn't wrong — AI systems do extract passages, and clean structure does help. But the overcorrection into content that's all structure and no substance was predictably doomed.

What I took away from the Milan event most is this: Google is trying to tell you that optimizing for AI search and optimizing for traditional search are not separate tracks. They are the same track. Write well, build authority, structure clearly, and don't try to pre-fragment your content into what you think machines want. The machines are sophisticated enough. The gap you're trying to exploit doesn't exist the way the GEO consultants say it does.

Focus on your site's overall quality. Fix the thin stuff. Add depth where you have none. That's the Milan message, and it's the right one.

What to Do Today

• Pull up your top 20 blog posts — how many are thin "chunked" pieces vs. genuinely in-depth? Make a list of which need to be rewritten or merged.
• Check your site-wide quality signal: if more than 30% of your indexed pages are thin or have very low engagement, that's a problem to solve before publishing more.
• Review any content you rewrote specifically for AI optimization in the past year. Is it still readable and useful to a human? Be honest.
• Make sure your robots.txt is consistent — if you're allowing Googlebot, you need a clear, intentional policy on AI crawlers rather than a hodgepodge of allow/disallow rules.
• Add proper Article and FAQ schema to your most important pages. This is a legitimate structural signal that helps AI context without harming readability.
• Check your AI search visibility — are you appearing in AI Overviews and AI Mode results for your key topics? If not, that's a separate diagnostic worth running.