Most “what Google wants” content is guesswork dressed up as insight. This one isn’t — it’s a fairly literal transcript of what Reid said publishers get wrong, and what she thinks would help. Worth treating as a checklist rather than a vibe. Below, every point from the interview, expanded.

1 Stop publishing the same format for everything

Format mismatch

Reid’s first complaint about publishers is structural, not about quality. She says many sites keep producing the same format — usually a written article — regardless of what the topic actually calls for. Some subjects are better explained in video. Some need a quick visual. A written article is the default because it’s the easiest thing to produce, not because it’s what the reader wants.

What this means practically: before you write your next piece, ask what format the topic naturally wants. A recipe wants a video or a structured list. A debugging walkthrough wants a screen recording. A philosophical argument wants long-form prose. Matching format to topic is a content decision, not a technical one — it costs you nothing extra to make it deliberately.

2 An AI Overview is not your competitor

Depth still has a job

Reid was explicit: an AI Overview is not meant to substitute for a real ten-minute article. It’s a summary layer, not a replacement product. If your content is getting “answered away” by an AI box, that’s often a sign your piece wasn’t doing anything an AI summary couldn’t do — not proof that long-form writing is dead.

3 Lean into what you’re actually good at

Specificity over coverage

This is the most repeated idea in the whole interview. Reid says the publishers who do well are the ones producing content where they genuinely have an edge, instead of trying to cover everything broadly. Generic coverage competes with everyone. A specific angle competes with almost no one.

She framed this as good news for small publishers specifically — calling it “an awesome time” for them, because uniqueness is now easier for Google to match to a specific searcher’s specific query, instead of small sites getting drowned out by bigger ones covering the same ground.

4 If you paywall, expect the traffic drop

No sympathy here

Reid wasn’t gentle on this point. Publishers who add a paywall and then see traffic fall are, in her words, just seeing what happens when you charge. Her proposed fix isn’t “don’t paywall” — it’s a future where Google routes a reader’s existing subscriptions to the publishers they already pay for, rather than surfacing gated content to people who can’t read it anyway.

If you do plan to monetize with a paywall, go in expecting a visibility cost, and make sure what’s behind it is valuable enough to justify losing some search/AI traffic in exchange for direct reader revenue.

5 Make sure Google can actually access your content

The boring, essential basics

Reid called Search Console the “webmaster console” in passing — a small but telling detail about how she still thinks about this from the crawling-and-indexing era. Her advice is unglamorous: confirm Google can technically reach and read your content before worrying about anything more advanced. No robots.txt blocks, no broken rendering, no content trapped behind JavaScript Google can’t execute, no accidental noindex tags.

6 Build an actual audience — it compounds

Loyalty signals carry weight

Reid’s repeated point: audience-building isn’t separate from SEO or AI visibility — it feeds directly into both. A reader who returns to your site, follows you, or marks you as a preferred source is a signal that compounds over time across search and AI Mode. There’s no shortcut around this; it has to be earned with real readers.

7 Google publishes actual guidelines — use them

Not guesswork

Worth knowing: Google has a published AI-optimization guide for site owners. Reid referenced it directly in the interview. If you’re going to optimize for AI visibility at all, start from Google’s own documentation rather than secondhand SEO folklore — it changes often enough that outdated advice circulates fast.

8 Search Console’s AI reports still have a gap

No click data, yet

Google has rolled out AI performance reports inside Search Console and recently expanded who can access them. But there’s still no click data for AI Mode or AI Overviews. Reid’s response, when pushed on this, was that the team is “still figuring out what metrics are useful” — essentially admitting the measurement layer hasn’t caught up to the product.

Her advice to publishers in the meantime: build your own analytics around what’s actually valuable to you — sales, signups, downloads, subscriber growth — instead of waiting for Google to hand you click numbers.

9 On filter bubbles and “preferred sources”

Her counter-argument

Reid pushed back on the idea that personalization creates filter bubbles that hide small publishers. Her reasoning: there’s no single “best shoe for everyone,” so some personalization is just realism, not bias. She also pointed out that Top Stories already blends preferred sources with non-personalized top sources — it isn’t all-or-nothing. AI Overviews, she said, try to give a topic overview rather than narrowing to one trusted voice.

Her own framing of the trade-off was candid: Google isn’t narrowing what information exists, but it’s also not telling people who they should trust. That’s a real tension she left unresolved — worth knowing she’s aware of it, even without a fix.

10 What Google is actually optimizing for

The metrics that matter to her

• Whether users are more satisfied with the search experience
• Whether users come back to Google more often
• Whether people bring more of their questions to Google — a trust signal
• What questions people aren’t asking yet, that Google could start serving
• Long-term: helping people actually understand and use information, not just access it

None of these are publisher-visibility metrics. They’re Google’s own internal success criteria. Useful to know precisely because they tell you Google’s incentives aren’t aligned with “send more traffic to sites” by default — traffic is a side effect of these goals, not the goal itself.

11 Even Google admits bugs slip through

A small but honest admission

Reid noted Google does extra testing for YMYL queries (Your Money, Your Life — health, finance, safety topics) but acknowledged bugs still get through. If you ever see something clearly wrong about your site in an AI Overview, it’s not necessarily intentional demotion — it might genuinely be a bug, and worth reporting through Search Console feedback channels rather than assuming malice.