Skip to content
ceeme

What does someone mean when they say AI SEO?

AI SEO is not a defined craft but an umbrella term the market uses for three incompatible things: having a language model generate content, making your website fit to appear in AI answers, and having software perform the SEO work itself. Anyone hearing the term without asking which of the three is meant stands a fair chance of buying something other than what they think.

By Njusja Orban, founder of ceeme3 min read

Measure your own company → Rather talk it through first

In short

Meaning one
Having AI write texts. That is a means of production and says nothing about being found.
Meaning two
Being findable inside AI. That is what is elsewhere called GEO or AEO, and the only meaning with a measurable outcome.
Meaning three
Automating SEO work. Useful, but it changes who does the work, not what gets measured.

Why does one term cover three different things?

Because the term comes from sales and not from the craft. Sticking "AI" in front makes an existing service new without changing anything about it, and that is attractive in a market where everyone must do something with AI at once. The result is that two quotes with the same title can cover completely different work.

Which question immediately exposes what is being sold?

This one: where is the result, and how do I check it? For the first meaning the answer is a number of pages written. For the third it is hours saved. Only the second gives an outcome you can verify yourself by asking the assistant the question again.

And that contains the second question: which questions get asked, and are they your customer's questions or your field's? A service measuring on category questions shows you exist among people who already know the word. That is the easy half.

Does content written by AI still work?

Yes, and it is not the origin that counts but what is in it. A text gets picked up because it holds a fact stated nowhere else as clearly, with something underneath backing it. A text that only repeats what is everywhere already does not get picked up — whether a human or a model wrote it.

So the practical risk of generation is not that it stands out, but that it is empty. Fifty pages that read nicely and assert nothing checkable are fifty pages nothing can be quoted from.

How do I recognise an offer with substance?

By three things that all belong in writing: which questions get measured, at which engines, and in which market. If one is missing the outcome cannot be repeated — and a measurement you cannot repeat cannot be laid beside the next one three months from now.

Frequently asked questions

Does a search engine penalise AI-written text?

The stated rule is about usefulness, not origin: text produced at scale without adding anything gets penalised, regardless of who typed it. That makes the question practical: does this page add something found nowhere else? If not, the risk exists without AI too.

Do you use AI yourselves to write these pages?

Yes, with one hard rule attached: every figure and every claim about what our system does comes from the code and is re-checked at every build. If the text does not match what runs, the build fails. That is exactly what we measure at clients, so it would be odd not to do it ourselves.

Read on

Free, no account and no card.