Google’s Liz Reid: AI Search Success Requires Reader-First Content

Reading time: 4 min

Publisher Traffic Isn’t Just Dying Because of AI

Liz Reid, VP of Search at Google, recently sat down to address the elephant in every publisher’s room: why traffic is tanking, and what to do about it. Her first response caught me off guard with its candor. She’s not blaming AI overviews entirely. Not even mostly, according to her.

She pointed to shifting user behavior. People are consuming information in different formats — notably, video and social platforms. She cited data from the Reuters Institute showing this migration away from traditional text-based search results. Here’s what actually happened: the consumption landscape has fragmented, and AI is just one piece of that puzzle.

Let me show you the data. The same Reuters Institute report she references shows a steady decline in direct traffic to publisher sites from Google Search starting as early as 2020 — pre-AI overviews. I’ve seen this play out before. Think back to 2012 when Google pushed authors and structured data. Many editors blamed that shift for killing their traffic. But really, it was the rise of Twitter and Facebook as news destinations.

Blaming AI alone misses the bigger pattern — audiences are diversifying their sources. For publishers, the uncomfortable question is whether text-only articles can still command attention.

Technical Crawl Access Is Your First Fight — Don’t Lose It

When the conversation turned to AI search visibility, Reid broke her answer into two parts. First: the technical basics. If your content blocks Google’s crawlers, your fight is over before it starts. She emphasized that Search Console provides controls for publishers to manage access — but blocking actually cuts off any chance of appearing in AI-generated answers.

I can sum up her exact words: “The first thing is making sure that we can access your content. If you block it, it won’t work. We provide tools in the Search Console so you can make choices … but facilitating access is step one.”

This sounds like SEO 101, and it is. But here’s the twist. Some publishers, frustrated with traffic losses, have started using robots.txt to block known AI crawlers from OpenAI, Anthropic, or even Google’s internal ones. In theory, that looks like protecting IP. In practice, Reid makes clear: blocking crawl is a visibility tax you don’t want to pay.

Search-Engine-First Content Is Dead, and Google Just Officially Buried It

The second part of Reid’s answer is harder to stomach for many publishers. She drew a clear line between content made to satisfy a search algorithm and content someone actually wants to read.

“You have to produce content people want to read,” she said. “If you produce content just for search engines … people are going to figure it out.”

Slow down. Think. This is not a new message, but the stakes are higher now. AI overviews can summarize the generic, the surface-level, and the “thousandth copy of the same story.” What they cannot replicate — yet — is original expertise, unique perspective, data you collected, experience you paid for. A review of every blender on Amazon might be a commodity. A review based on six months of daily testing at a restaurant supply store? That’s a different thing entirely.

She pointed to Google’s updated guidelines for creators. The metrics that matter: freshness, relevance to real-time queries, and depth that can’t be found in fifty competing articles. The mechanism is simple. If users click a result and actually read it — time on page, scroll depth, no pogo-sticking — that’s a signal Google’s system amplifies. If users bounce, nothing else can fix it.

The Real Worry: Quality Publishers Who Are Still Losing Traffic

Reid’s diagnoses feel logically sound. But they leave a gap large enough for a truck to drive through. She’s describing what should work, not necessarily what does work for everyone.

I’ve talked with independent publishers who run sites focused on knitwear, niche electronics repair, and hyper-local real estate data. They follow every guideline Reid outlines: original research, expert authorship, no thin content, high session duration. Some of them have seen organic traffic drop by 30–40% in the last year. Nobody talks about this part of the story.

This isn’t a take — it’s a pattern. Large media operations with hundreds of writers and AI-generated content farms have been hit hardest, yes. But the algorithm doesn’t always distinguish well between “quality but narrow” and “thin.” The risk is that AI overviews treat a 2,000-word expert guide and a 200-word rephrasing of a press release as equally suitable for summarization — when both contain similar surface-level claims.

What This Actually Means for Your Publishing Strategy

Reid’s interview doesn’t upend SEO fundamentals. It reframes them. AI search doesn’t reward generic content because generic content gets eaten alive by the summary box. The only content that survives the summarization filter is the stuff that adds context, data, perspective, or actionability the AI can’t extract from a Wikidata entry or a raw press release.

For publishers: guides, product tests with hands-on experience, investigative journalism, essays with strong points of view — these have a defensible position. But every piece needs a reason to exist beyond targeting a short-tail keyword. Every article needs to ask itself: “Does this bring something a machine can’t?”

The playbook changed. Again. But the direction is coherent: write for humans, structure for machines, and measure by engagement. If you do those three things, you’re playing the right game — even if the scoreboard looks different than it did in 2024.

Matt S. Riley

Building websites since before Google existed. I’ve run SEO, growth, and content for startups across California — and I’ve watched every ‘revolutionary’ tactic eventually expire. What doesn’t expire: understanding systems, compounding effort, and thinking slower than everyone else.