The link prediction model: how internal links steer your AI visibility

A link prediction model predicts where links naturally belong in a piece of text.

The verdict: internal links survive AI extraction when they sit where an editor would put them. Forced or promotional links get stripped.

You write a carefully crafted paragraph with an internal link to your product page or a related article.

An AI system extracts that paragraph and uses it in an answer. And the link? Gone.

Another time the link does survive. It shows up as a clickable path to your site, right in the middle of the generated answer.

That difference isn’t random. It follows patterns you can understand and steer.

The link prediction model is a way to predict where links naturally belong, so your internal links survive extraction by AI systems.

Below we explain how AI judges links, and what that means for the way you link internally.

This is part of our broader view on generative engine optimisation (GEO), worked out in full in the ultimate GEO guide.

Check your own page: get a GEO score with concrete tips through our free GEO check.

Why do links in AI answers matter?

Because a surviving link is one of the few moments someone actually lands on your site.

In classic search results, every result is a link. In AI answers, that’s not the case.

The synthesis often answers the question without the user clicking anything.

Four reasons a surviving link counts

But it’s about more than clicks:

• Direct traffic: a surviving link is a direct path to your page, in an environment where that traffic has become scarce.
• Brand visibility: even without a click, your domain shown as a clickable source signals authority and relevance.
• Attribution: the link credits you as the source. Without a link, your contribution becomes invisible.
• Compounding effects: visible domains build familiarity, and link patterns can influence future selection.

Because direct traffic from AI searches is limited, every surviving link becomes disproportionately valuable.

Our guide to citation rate shows how to measure how often you’re cited.

How do AI systems handle links?

They decide per link: keep it, turn it into plain text, or move it to a source list.

When an AI extracts content, it encounters links as HTML elements. Then it has to make a choice.

The four factors that weigh in

The factors look a lot like editorial common sense:

• Relevance to the answer: does the link fit what the AI is currently explaining?
• Natural placement: does the link sit where a link would normally sit, or does it feel forced?
• Authority of the source: if the system trusts the source, it’ll probably trust its links too.
• Value to the user: does keeping this link add anything, or is it pure self-promotion?

What happens to links that don’t survive

A link doesn’t always disappear entirely. Often it gets transformed:

Knowing these transformations helps you set more realistic expectations and target your optimisation more precisely.

What citation conversion looks like

Here’s citation conversion in a real answer. Perplexity turned its sources into small chips behind each bullet, plus a separate sources panel:

Perplexity, our own test prompt, 17 October 2026: inline source chips and a “10 bronnen” (10 sources) button instead of links inside the text (Dutch UI).

The page still gets credit: Perplexity lists 10 sources for 3 bullets. But the link no longer sits inside the sentence.

When links vanish completely

Some answers drop links altogether. The free ChatGPT answer to a similar prompt listed no sources at all:

ChatGPT free (temporary chat), 17 October 2026: three bullets, about 90 words, no sources.

So placement isn’t the only factor. The engine and its settings decide too.

What is the natural link pattern?

The way links appear in quality editorial content, which AI systems learn from.

Here’s the core insight: AI systems learn link patterns from their training data.

That data contains millions of web pages full of links. From it, the model learns where links naturally appear: in which contexts, at which sentence positions, with which anchor text.

The question the model asks

When it decides during extraction whether a link stays, it’s implicitly asking one question:

“Does this link sit where links normally sit in quality content?”

Links that match the natural pattern survive more often. Links that feel forced get stripped.

Four recurring characteristics

Analysis of editorial content from quality publications shows a few recurring characteristics:

• Contextual relevance: the link sits exactly where a reader would want more detail, not as a detached call to action.
• Specific anchor text: descriptive and concrete (“real-time streaming architecture”) instead of vague (“click here”).
• First-reference linking: only the first mention of a concept gets a link, later mentions stay plain text.
• Information-dense context: links belong in paragraphs where the reader is actively learning something, not in an introduction or conclusion.

This links directly to how you structure your pages for extraction.

More on that in content architecture for AI extraction.

How does a link prediction model work?

It’s trained on editorial links, then scores every possible link position in new text.

These patterns aren’t just observations. They can be formalised into a prediction model.

Where the idea comes from

Link prediction is an established problem in network science.

Wikipedia defines it as “the problem of predicting the existence of a link between two entities in a network” (Wikipedia on link prediction).

Applied to content, the “network” is your text and the entities are the concepts you could link.

The training logic

A link prediction model takes text as input and predicts where links would naturally appear. The training logic looks roughly like this:

1. Collect content with natural editorial links (news sites, quality publications).
2. Record the context for each link: surrounding text, sentence position, anchor text.
3. Train a model to predict link presence based on those features.
4. Have the trained model score new content.

The patterns it picks up

The model picks up patterns such as:

• links often follow phrases like “read more about” or “according to”
• links rarely sit in the first or last sentence of a paragraph
• anchor text is usually 2 to 5 words long
• links mostly accompany the first mention of an entity

Reading the prediction scores

For every possible link position, the model gives a probability score:

• High probability (>0.7): this position strongly matches natural patterns.
• Medium probability (0.4 to 0.7): plausible, but not optimal.
• Low probability (<0.4): this position matches no natural pattern.

Those same thresholds can be drawn as an ascending scale. The higher the score, the more strongly the position follows the natural pattern, and the greater the chance your link survives:

What you do with the scores

The optimisation is simple in principle: move links from low to high probability.

You can also use the model in three other ways:

• test content before publication
• have optimal positions suggested
• analyse competitors

For example, say a paragraph ends with “Want to know more? Click here.” That link scores low.

Move it to the first mention of the concept, with the concept as anchor, and it scores high.

What are the five principles to optimise your link placement?

In practice it comes down to a handful of concrete rules.

1. Link at natural reference points

Place links where your content mentions something that invites elaboration:

• the first mention of an entity
• the introduction of a concept
• a source or figure
• a recommendation

Not as a detached “read more”.

2. Use descriptive anchor text

Replace “click here for pricing” with something like “see the pricing and plans”.

Google’s link guidelines put it plainly: good anchor text is “descriptive, reasonably concise, and relevant”. “Click here” and “Read more” are its examples of what to avoid.

The anchor text describes the destination and helps the AI assess relevance. Our guide to anchor text for internal links has more examples.

3. Link only the first mention

If an entity appears several times, link only the first substantial mention.

Repeatedly linking the same destination adds clutter for the reader and no extra context for the model.

4. Embed links in substantial content

A link in an information-dense paragraph is more natural than a link in thin text like “Want to know more? Check out our resources.”

Deliver value around the link.

5. Respect editorial density

Google says there’s no magical ideal number of links per page. But it does warn against stacking links right next to each other, because readers can no longer tell them apart.

Rule of thumb: a few well-chosen links per section is plenty.

One well-placed link survives more readily than seven competing links in the same paragraph.

The context around the link counts

The words right before and after a link influence its odds of survival.

Strong pre-link signals create the expectation of a reference. Weak signals betray promotional intent. Use this table:

Mid-sentence matches the editorial norm.

The same idea drives grounding snippets that AI likes to pick up.

How do internal and external links differ?

AI systems don’t necessarily treat the two equally.

Internal links

Internal links (to your own domain) remain an SEO fundamental.

Google lists making content findable through internal links among the basics that still matter for AI Overviews and AI Mode.

In answers, though, they’re sometimes seen as navigation rather than citation. They carry a lower authority signal, because you’re essentially pointing at yourself.

They stay valuable if the linked page genuinely adds something. Keep the anchor text descriptive and avoid over-linking that resembles a navigation menu.

Our guide to internal links covers the SEO basics; internal linking for AI covers the AI side.

See your own internal links

Search Console shows how many internal links each page gets. Here’s ours:

The Links report in our own Search Console for customerimpact.be, October 2026: 2,116 external links and 364,879 internal links.

That’s 364,879 internal links against 2,116 external ones. Template links such as the author page, with 3,584 links, lead the list. That’s navigation; the editorial links inside your articles are the ones AI judges.

Watch out: pages with almost no internal links are hard to find for crawlers too. Our guide to orphan pages shows how to spot them.

External links

External links (to other domains) carry a higher authority signal. They read as an editorial reference to a third party.

The risk is sending users away, so balance them with internal links.

Combine the two

One powerful pattern combines the two: an external citation of an authoritative source, followed by an internal link to your own relevant page.

That mirrors editorial practice and can improve the survival odds of both.

What is the link-content relationship?

Links don’t exist separately from the content they sit in.

The most important thing to remember: if your content is strong enough to be selected as a primary source, your links come along into the selection.

If your content isn’t selected, all link optimisation is wasted effort.

Quality first, placement second

So the order is clear:

1. fix the quality and the selection odds of your content
2. optimise the links only after that

Improving those selection odds is a discipline in itself, worked out under Selection Rate Optimization.

Never trade readability for a link

And never sacrifice readability for a linking opportunity.

The best link optimisation is invisible: the link sits where it belongs and adds value. It doesn’t disrupt the flow.

Want to see whether AI engines mention you at all? Run our free AI visibility checker.

Frequently asked questions

What exactly is a link prediction model?

It is a model that takes text as input and predicts where links would naturally appear. It is trained on large volumes of quality content with editorial links and learns from that which contexts, sentence positions and anchor texts correlate with links. You use it to test before publication whether your link placement feels natural, and to move links to positions with a higher probability score.

Why do some of my internal links disappear in AI answers?

Because during extraction the AI judges whether a link sits in a natural position. Links that feel forced, promotional or navigational are often stripped or converted to plain text or a standalone source reference. Links that match the editorial patterns, with descriptive anchor text in an information-dense spot, survive considerably more often.

How many internal links may a page have?

There is no fixed number: Google states there is no ideal link count per page. Too many links dilute the value of each individual link and send a downward quality signal. One well-placed link in substantial context survives extraction more readily than a handful of competing links in the same paragraph.

Should I optimise my content or my links first?

Content first. If your page is not selected as a source, your links get no exposure anyway, however well placed they are. Strong content that is chosen as a primary source carries the accompanying links into the selection with it. Link optimisation is the refinement, content quality is the foundation.

Can I automate internal linking safely?

Yes, if the tool follows the same rules: first mention only, descriptive anchors, mid-sentence placement. Our guide to automating internal linking explains where automation helps and where a human check is still needed.

Need help?

Want to translate this into execution? See how we approach this with AI findability.

We’ll review where your links sit today and which ones AI engines are likely to keep.