What we read this month, and what we changed

Two pieces of research crossed my desk this month that say opposite things. One says the ceiling on content quality has gone. The other says the floor on content integrity has fallen through. Both of them matter for anyone trying to work out how AI search actually decides what to recommend.

Here is what they say, and what we did about it.

The offensive case: quality is the floor, not the ceiling

Dan Taylor published a piece in Search Engine Journal a few weeks back that has been sitting with me since. His argument is sharp and uncomfortable.

Quality content used to be enough. You wrote something good, you ranked, you won. Taylor’s point is that quality is now the floor, not the ceiling. Everyone has good content. The models are not short of well-written, well-structured pages to pull from. What separates the brands that get cited from the ones that do not is distribution.

Not distribution in the old social-media sense. Distribution in the retrieval sense. Where does your content sit in the network? How many third-party sources reference it? How findable is it across multiple retrieval paths? Taylor’s framing: the unit of competition has shifted from pages to fragments, from ownership to retrievability. A network of average-but-distributed content beats one exceptional-but-isolated piece.

That is a hard thing for content teams to hear. It means the best article in the world, sitting on your own domain with no external validation, will lose to a decent article that five credible sources link to, reference, or build on.

For anyone measuring AI visibility, this reframes the question. It is not “is our content good enough?” It is “is our content findable enough, across enough surfaces, with enough third-party signal, that the model picks it up when someone asks a relevant question?”

The answer, for most brands we measure, is no. Not because the content is bad. Because the distribution is not there.

The defensive case: anyone can trick the system in 20 minutes

Then Thomas Germain at the BBC published the other side of the coin.

Germain ran a simple experiment. He wrote a blog post claiming to be a world-champion competitive hot-dog eater. Within 20 minutes, both Google and ChatGPT were citing the claim as fact. He is not a competitive eater. He is a journalist who wanted to see how easy it was to manipulate AI search results.

It was very easy.

Lily Ray, one of the most respected names in SEO, called it proof that AI search systems are not yet equipped to distinguish real authority from manufactured authority. Harpreet Chatha made a similar point: the models are pulling from the open web, and the open web is full of content that looks credible but is not.

Google has since quietly updated its spam policies to cover AI-generated misinformation. That tells you something about how seriously they are taking the problem.

For brands, the defensive case is just as important as the offensive one. If someone can become a world-champion hot-dog eater in 20 minutes, they can also become a credible competitor in your category. A fake review site, a handful of planted mentions, a Wikipedia edit. The signal architecture that AI models use to decide who to cite is not yet strong enough to catch all of this.

This is where measurement matters. If you are not monitoring what AI is saying about your brand and your competitors, you do not know whether the citations are accurate. You do not know whether someone has poisoned the well. You do not know whether the answer the model gives about you is based on real authority or on something someone put there last Tuesday.

What we changed

Zero.

No product changes. No new features. No methodology updates. The AVS report measures the same things it measured last month, in the same way, across the same engines.

What changed is the framing.

Taylor’s piece sharpened how we talk about distribution as a citation signal. We have always measured third-party presence. Now we have a better way to explain why it matters: because quality is the floor, and distribution is what gets you above it.

Germain’s piece sharpened how we talk about the defensive value of measurement. We have always said that knowing where you stand is the first step. Now we have a concrete example of how fast the landscape can be manipulated, and why regular monitoring is not optional.

The product does not change every time a paper comes out. But the thinking does. And the thinking is what clients pay for.

Focus. Measure. Plan. Deliver. Repeat.

Be Known. Be Cited.