Blog
Search3 min read

Being findable when the answer is written for you

Search is splitting in two. On what a peer-reviewed study found actually gets a page cited inside an AI-generated answer, and why it is not what the SEO industry is selling.

Rows of white archive boxes on wooden shelves in a panelled room.
SearchPhotograph via Unsplash

For twenty-five years, being findable meant ranking. Ten links, and the work was to be one of them. That has not stopped mattering, but a growing share of buying research now happens inside an assistant that reads the sources and writes an answer, and in that world you are not competing for a position. You are competing to be the thing the answer is built out of.

This has a name now, Generative Engine Optimisation, and an industry has appeared around it faster than any evidence has. Almost everything published about it comes from firms selling the service. There is one exception, and it is worth more than the rest put together.

The study

Researchers from Princeton, Georgia Tech, the Allen Institute and IIT Delhi built a benchmark of 10,000 real queries, tested nine different ways of changing a page, and measured which ones made a generative engine more likely to cite that page in its answer. It was published at KDD 2024, one of the two serious venues in the field, and peer reviewed.

40%
relative improvement in a source's visibility inside generative engine answers, from the top-performing content changes.10,000 queries across multiple domains · Aggarwal et al., GEO: Generative Engine Optimization, ACM KDD 2024

Then they tested it against a real, live generative engine rather than a laboratory setup, and it held.

37%
visibility improvement measured on a real-world generative engine, not a simulation.Live test · Aggarwal et al., ACM KDD 2024

What actually worked

Nine strategies were tested. Three of them moved the needle, and they are not the three the industry is selling.

  • Citing sources. Attributing claims to the organisation that published them.
  • Quoting credible sources directly rather than paraphrasing them.
  • Adding statistics. Replacing qualitative discussion with quantitative figures wherever the page can honestly carry them.

Those three achieved a 30 to 40% relative improvement. Keyword density did not appear. Neither did anything resembling the tactical work that dominates SEO advice.

The things that make a page more likely to be cited by a machine are the things that make it more trustworthy to a person. That is not a coincidence, and it will not last forever, but it is true now.

Why this is not surprising once you think about it

A generative engine is deciding which sources to build an answer from and which to attribute. It is doing, badly and at speed, what a researcher does: preferring material that says where its claims came from, because that material is easier to justify repeating.

A page that says “studies show engagement improves” gives an engine nothing to stand on. A page that states a specific figure, names the organisation that published it, gives the period it covers and links to it, gives the engine a citable object. The first page is unquotable. The second is the answer.

What this means for the classic half

Ranking has not gone away, and the discipline underneath it has not changed: a page has to be fast, legible to a machine, and about something. What has changed is that being described consistently everywhere you appear now matters as much as being optimised in one place, because an assistant assembling an answer about you is reading all of it at once and has no way to resolve a contradiction in your favour.

  • Say the same thing about yourself everywhere. One description, one set of facts, one name.
  • Publish structured data. It is the difference between being read and being parsed.
  • Put a real figure with a real source on any claim worth making. It is now the highest-leverage thing on the page for both audiences.
  • Do not buy a GEO package priced against a promise nobody can measure. Ask what evidence it rests on. There is exactly one honest answer and it is a paper.

This site is built that way, which is not a coincidence either. Every figure on it carries the organisation that published it, the period it covers and a link. That was a decision about honesty made before any of this research was read. It turns out to also be the thing that gets you quoted.

Sources

01

New writing, when there is some

A short note when something is published: what it is about and a link. No more than a couple a month, and nothing else.

Start with
the diagnostic.

A scoping conversation costs nothing and ends with a straight answer about whether there is enough here to be worth doing. If there is not, we will tell you.

Book a scoping callinfo@skaiyon.com

Questions. Asked before every engagement.

Your data is yours. You can export it at any time, and it is exported to you in a documented format before any engagement closes. The software itself is licensed: we build it around your business, host it, and run it, and you pay monthly for that. If you'd rather own it outright, that's possible. It's a different kind of engagement and it's priced accordingly. Licensing keeps maintenance our problem rather than yours, which is why it is the default.

Two weeks' notice, either side. Your data is yours. You can export it at any time, and it is exported to you in a documented format before any engagement closes. The system stops running. If you'd rather keep it running, the ownership option is available at that point as well. Ending the retainer doesn't force you to lose what was built.

Typically four to eight weeks from signing to a working system, depending on how many tools it connects to and what state the data is in. A scoping conversation and a written plan come first, so the timeline is agreed before anything is committed to.

Access to the tools and data the system will work with, one person who can make decisions, and a few hours in the first two weeks while we map how work actually moves through your business. After that, very little. The point of the engagement is that it runs without your attention.

Usually. Most business software exposes an interface we can build against, and where one doesn't there's normally a way around it. Which connections are viable is settled in the scoping conversation, so you find out before committing rather than after.

Hosting region for a system we build for you is decided at the start of the engagement, not inherited from a default. If data residency is a hard requirement, raise it in the first conversation and we will tell you plainly whether we can meet it. This website is separate and its processing is already fixed: enquiries, quote requests, bookings and chat are handled by Supabase in Tokyo, Resend in Tokyo, Cloudflare, cal.com and Moonshot AI, which the privacy policy names individually along with where each one processes. Nothing from a client system is ever sent to the chat assistant.

That's the normal starting point, and mapping them is part of the work. Automating a process nobody has examined just makes the confusion faster, so we don't start there. The first phase establishes how things actually happen, as opposed to how they're supposed to.