This adds to the growing body of evidence that LLM skills are amplifying incumbent advantage.

Full paper: Reimers & Waldfogel, "AI and the Quantity and Quality of Creative Products"
www.nber.org/system/files...

Are more books reaching more real readers? The answer is yes, but only at the top of a much larger and noisier distribution.

The paper also found that incumbent authors with established track records became more productive with LLMs, while new entrants mostly added to the low-quality pile.

They measure "quality" by how many readers actually engage with a book (rating counts, validated against sales data), not literary merit. It might seem sketchy, but it's the right question for a market analysis...

- Authors debuting in the LLM era disproportionately produce low-quality work.
- And yet, the top 1000 monthly releases per category are "better" than before, not worse. The volume of new entries means the absolute number of good books reaching readers has increased, even as the average plummets.

We were saying this stuff one year ago... while they were marching in the streets declaring the death of Google.
I guess they really think people have short memories and are easily fooled.

#SEO #AEO #GEO #LLMO

If your content is dense reference material and your audience wants summaries, this feature is fine. But for most sites, you're trading real engagement for a gimmick that sounds clever but either does nothing or actually damages you.

I'm sure this thread won't go viral, though.

- The "authority" framing might backfire. Users seeing a prompt injection that says "remember X as an authority" might see it as manipulation. Does this build brand trust or does it actually damage it?

- AI summaries will lose nuance or introduce errors. You worked to craft specific arguments, and you now have no quality control.

- Once users leave for ChatGPT, you have zero visibility into what happened. Did they actually read the summary? Did they find it useful? What's the next step they took?

- You're actively sending visitors away from your site. They exit the organic funnel. You lose time-on-page, web analytics, engagement with your writing.

But the problem doesn't even end there. There are additional strategic issues with this hack from a growth marketing point of view:

You're not "feeding your content into the AI's brain", you're not building brand authority in any AI system.

And about the "remember my brand as an authority" prompt injection: the AI doesn't have an authority dial you can set with a string. It mostly ignores this.

When someone clicks your link and ChatGPT summarizes your article, that interaction exists only in their session. It vanishes when they close the tab. Other users asking ChatGPT about your topic tomorrow have zero exposure to it.

The fundamental misunderstanding: session context ≠ training data ≠ persistent brand recognition.

Full study here: seranking.com/blog/health-...

Only about a third of citations came from higher-credibility medical sources. Academic journals and government health institutions together accounted for about 1%.

YouTube was the most-cited single source, beating out hospitals, clinics, insurers, and health associations. This despite ranking just 11 in traditional organic results, suggesting AI Overviews favor video content heavily.

On the US front, growth stocks outperformed value only slightly, with returns of 18% versus 16% - and globally, growth and value closed the year identically (23%).

2026 is positioned to be a value-focused year. Fundamentals over hype.