Ask any newsroom editor how they feel about turning a story into a string of micro-paragraphs and you’ll get a laugh — and now Google wants publishers to stop doing it for search engines, too.
On the Search Off the Record podcast (around the 17–18 minute mark), former Search Liaison Danny Sullivan was blunt: don’t rework your content into “bite-sized chunks” because you think large language models will reward it. Sullivan said he checked with Google engineers and repeated the company’s long-standing refrain: write for people, not for ranking systems.
What Sullivan meant by “chunking”
If you’ve been online in the past year you’ve probably seen the pattern: pages made up of very short paragraphs, frequent subheads formatted as user questions, and a structure that looks more like a chatbot prompt than a human-readable article. The logic behind it is straightforward — split content into small, extractable units and hope AI systems ingest and surface them as clean answers.
Sullivan and other Google search advocates, like John Mueller, argue that this is a misguided adaptation to a temporary behavior of some systems. It might produce a blip in traffic now, but it’s not a durable strategy.
Why Google thinks chunking is a bad bet
There are three practical objections. First, Google’s ranking systems are designed to reward content that serves human readers — and those systems will continue to evolve that way. Second, building two versions of the same content (one for humans, one for machines) is expensive and brittle. Third, short-term gains from exploiting today’s quirks can evaporate when models and ranking signals shift.
Sullivan acknowledged edge cases — yes, some publishers may see small wins — but asked whether reshaping your content, reorganizing teams, or creating separate publication streams is worth the risk when the systems change.
Short-term wins, long-term risk
SEO has a long history of trend-chasing. Whenever a new feature or signal pops up, some sites try to optimize solely for it. Sometimes it works for a while. Often it doesn’t last. Sullivan’s point is practical: if you spend marketing and editorial energy trying to outsmart the current behavior of LLMs, you could end up with content that’s less useful to real people and less resilient when algorithms evolve.
This guidance comes as Google extends AI into more products — from conversational search experiences to deeper Workspace integrations. For context, Google’s broader AI rollouts (for example, features tied to Gemini and new AI modes across services) show the company is still figuring out how models should work alongside traditional search interfaces. See how Gemini Deep Research and new AI Mode features are being integrated into Google’s ecosystem for one example of that evolution.
What publishers should do instead
The advice isn’t glamorous, but it’s steady: focus on real readers. A few practical points:
- Prioritize clarity and usefulness. If a visitor finds the answer quickly and takes the next logical step (reads further, signs up, contacts you), that behavior matters to ranking systems.
- Invest in subject-matter quality rather than formatting tricks. Deep, well-sourced explanations tend to persistently attract links, citations, and repeat visits.
- Treat AI features as another distribution channel, not a reason to rewrite your editorial identity. Convertible short summaries or TL;DR boxes are fine when they genuinely help users — not when they’re only there to game an LLM.
Sullivan used a plain example: a local plumber should spend more time getting customers and good reviews than restructuring web pages into machine-friendly snippets. The same applies to small businesses and publishers: align content with business goals, not ephemeral ranking hacks.
A reminder for marketers and tool vendors
Google’s team also urged skepticism toward services that claim to speak for Google. If an agency or tool says “Google wants this,” ask for a citation. The company won’t endorse specific third-party products, and many claims in the SEO ecosystem are extrapolations rather than official directives.
The conversation on the podcast is part of a broader pattern: Google consistently nudges creators back toward the fundamentals of useful content. That might feel conservative in an era of rapid AI change, but it’s an argument for durability over opportunism.
Whether you’re a newsroom editor, a small-business owner, or an SEO specialist, the simplest test remains useful: would a real person read this and get value from it? If the answer leans toward yes, you’re probably building something that will weather the next model update. If not, you might be writing for a snapshot of today’s system behavior — and snapshots fade.