Our reading
Why this paper matters more than the content around it
There is an enormous amount of confident writing about how to rank in AI answers, and nearly all of it is vendor content with a product attached. This is the paper that defined the term, it was peer-reviewed at a serious venue, and it ran actual experiments.
That does not make it correct forever. It makes it the only place to start that is not marketing.
The finding with the most bite
Keyword stuffing did not work. The single most transferred habit from the SEO era was among the things they tested, and it did nothing.
What did work reads like a description of a decent source: quote people, include real numbers, cite where things came from. The plausible reason is structural — a language model assembling an answer needs quotable, attributable material, and a page dense with keywords and thin on substance offers nothing to lift.
What we take from it
Write things worth quoting. Not phrasings tuned for a model, but sentences that state something specific enough to be extracted and stand alone. The overlap with good editing is nearly total.
Put numbers in. A page that says "engagement rates are falling" is unusable. A page that says which rates, measured across what, over what period, is quotable — and checkable, which is the same property.
Cite your sources. It helped in the study, and it is the right thing to do regardless. This section of our site is partly an argument that those two facts are usually the same fact.
Expect it to shift. Anything tuned this closely to a particular model's behaviour has a short shelf life. Habits that make content genuinely better survive model updates. Tricks do not.
The part nobody selling GEO will tell you
Being cited is not being visited. A generative engine that quotes your page accurately may have removed the reader's reason to arrive at it. There are second-order arguments for the citation anyway — being the named source, in front of the right person, at the moment they ask — and we think those arguments hold. But if someone quotes this paper's 40 percent at you as a traffic projection, they have either not read it or are counting on you not having.
We sell this service. That is exactly why the caveat is on the page.
The study above is the work of its authors and is not ours. The summary and commentary on this page are written by Big Bang Story and are our interpretation, not the authors’. We do not host copies of other people’s papers — read it at the source.