AI SEO: how you get mentioned in ChatGPT answers

Three pillars determine whether ChatGPT mentions you: citable structure, demonstrable authority, and presence where AI systems look.

← back to home

AI visibility // approach
ai-seo.md

$ build --field=ai-seo --pillars=structure,authority,presence

AI SEO: how you get mentioned in ChatGPT answers

AI SEO is the discipline of getting mentioned in AI answers, and it requires three things: content with a structure that can be cited, authority in the form of sources and figures, and presence in the places AI systems look. There are no guarantees, but each of the three can demonstrably be improved and measured. This piece describes the approach; you can do most of it yourself.

DD DataDrift Digital 14 September 2026 6 min

ChatGPT assembles its answers from texts it can read and cite. So the question isn't how you put the model in a favourable mood, but how you make material it can actually use. That work is now called AI SEO, it breaks down into three pillars, and all three are simply work rather than magic. Anyone who has all three in order needn't do anything special beyond that; anyone who skips one keeps wondering why the competitor gets mentioned.

PillarWhat it requiresFirst action
StructurePassages that answer a question on their ownPutting the answer at the front on your key pages
AuthorityFigures with a source, a named authorBacking up claims and putting a name under your pieces
PresenceAccurate mentions in the places where your market is describedChecking and updating existing mentions

Which pillar deserves priority for you is told by the measurement: so start with a baseline measurement, not with building.

01 / structureMake your content citable

AI systems don't cite websites but passages: standalone pieces of text that answer a question on their own. The first pillar is therefore that your pages consist of such passages.

AI doesn't cite websites. AI cites passages.

Two interventions do most of the work here. Put the answer at the front: the question in the heading, the full answer in the first paragraph, and only then the explanation. And give each piece a form that can be parsed: headings that match real search queries, a table where something is compared, a list where something goes step by step. The full writing rules are in how to write content AI can cite.

The good news for smaller players: this is the pillar where the established names regularly drop the ball. Many large sites carry years of stacked marketing copy in which the answer is never at the front. One well-structured page about one genuine customer question can therefore be chosen over a domain that is, on paper, much stronger.

02 / authorityGive the model a reason to choose you

Between ten pages claiming the same thing, an AI system picks the page that backs up its claims. Princeton's GEO research (KDD 2024) measured this: source citations increase visibility in AI answers by around 40 percent, statistics by 37 percent and expert citations by 30 percent.

Here too, two interventions are the starting point. Back up your claims with figures and name the source along with the year, even when that figure isn't your own. And put a name under your pieces: a named author with a recognisable background carries more weight than an anonymous company blog. What you leave out counts too: inflated claims without a source actually make a page less citable.

03 / presenceBe where the model looks

ChatGPT doesn't draw its picture of a market only from company sites, but mainly from the places where that market is described: trade media, review sites, comparisons, forums. Having your own site in order is therefore half the work; the other half is being present in the sources about your field.

Concretely, that means: make sure mentions of your company on third-party sites are accurate and up to date, and genuinely take part where your field is discussed. Fake mentions and bought praise get found out and have no place here. Why those third-party sources carry so much weight is covered in why AI refers to third parties more often than to your site.

04 / the orderWhere you start

Start with measuring, not building. Fix a set of search queries, note where you do and don't appear now, and you'll know which of the three pillars is weakest for you. Without that baseline measurement, every improvement is invisible and every investment is a gamble.

After that, the logical order is: first the structure of your key pages, because that's work you do once and it stays in place. Then the backing up, piece by piece. Presence with third parties runs through it all as the slow work: it builds up over months and can't be forced.

Keep the pace realistic. Five pages that genuinely answer the key customer questions do more than fifty written just to be there. And don't expect movement within a few weeks: indexing, re-crawling and the updating of answers take months, and that's the same for everyone, not just you.

05 / the honestyWhat no one can promise you

Positions in AI answers cannot be guaranteed, by anyone. Answers differ by session and by platform, and the selection mechanism belongs to the model builder, not to you or an agency. What can be done: carry out the method, record the measurement and show the movement in black and white. Why that's the only honest offer in this field is covered in can positions in AI answers be guaranteed. Anyone who promises you something else is selling something they don't own.

Frequently asked questions
What determines whether ChatGPT mentions your company?+
Three things together: whether your content has a structure that can be cited, whether your claims are backed up with sources and figures, and whether you're present on the third-party sites where your market is described. You can work on all three, and all three are measurable.
What does backing up content demonstrably deliver?+
Princeton's GEO research (KDD 2024) measured the effect of interventions on visibility in AI answers: source citations around plus 40 percent, statistics plus 37 percent and expert citations plus 30 percent. Backing up claims is therefore the best-documented intervention that's within your own control.
Where do you start if you're not mentioned anywhere?+
With a baseline measurement: a fixed set of search queries, noting per query whether you appear and who does. Then first tackle the structure of your key pages, then the backing up, and in parallel bring the third-party mentions into order. Without a baseline measurement, every improvement stays invisible.
Does it help to cram lots of keywords into your text?+
No, it works the opposite way. The same Princeton research (KDD 2024) found that keyword stuffing lowers visibility in AI answers by around 10 percent. AI systems choose readable, well-backed passages, not text that's been stuffed for an algorithm.
Can an agency guarantee a mention in ChatGPT?+
No. The selection mechanism belongs to the model builder and answers differ by session, so no one can guarantee a position. What a serious party can guarantee: a fixed measurement method, an upfront baseline measurement and monthly re-measurements that show the movement in black and white.
baseline measurement · fixed price · no obligations

Want to know which of the three pillars is weakest for you?

The AI visibility scan

We measure using a fixed set of search queries where you do and don't appear in ChatGPT, Perplexity, Claude and Google AI Overviews, and who currently does. One report, €295, fully offset against Op de Kaart. Even without a follow-up, you keep the report and know where you stand.

→ Request the scan

This text was produced with AI assistance and checked and approved for publication by a human.