AI does not rank companies quietly anymore.
It replaces them.
When someone asks ChatGPT for alternatives, it doesn’t browse.
It doesn’t compare ten links.
It decides. Simple.
This edition is about the fastest way to make sure that decision includes you.
Search trained us to chase positions.
AI forces a different question: When the leader disappears from the answer, who takes their place?
ChatGPT does not retrieve results. It constructs responses.
And those responses have structure:
A primary option.
A recommended alternative (sometimes two).
And that recommendation is not driven by who published more comparison pages or who ranks first on Google.
It is driven by whether the model already understands you as belonging in the same category and role as the leader.
This is where most teams are still operating with the wrong mental model.
SEO was about visibility.
AI is about familiarity.
If the model does not already recognize you as a legitimate substitute, you do not get considered. Even if your site outranks everyone.
The experiment that changed how we approach alternatives
We wanted one of our clients, let’s call them X, to get recommended for queries like: “best Y alternative”.
The obvious play was predictable → Publish multiple “best Y alternatives” articles. And push X as the first recommendation everywhere.
That approach works in Google. BUT… Inside ChatGPT, it didn’t.
So we reversed the strategy.
Instead of flooding the model with claims about replacing Y, we published just three articles.
Two of them were framed as “best X alternatives”, where Y was placed as the first recommendation.
Only after that did we publish one article titled “best Y alternatives”, with X positioned at the top.
No scale. No repetition. No aggressive self-assertion.
Yet this sequence fed more signal into LLMs than dozens of traditional comparison pages ever could.
Because the model was not learning that X wanted to replace Y. It was learning that X and Y belonged in the same category.
That distinction is everything.
The fastest way to become the recommended alternative
Here’s the principle most teams miss.
ChatGPT does not recommend alternatives because you try to replace someone. It recommends alternatives because it has already learned that you sit in the same mental bucket.
LLMs do not learn through dominance. They learn through association.
When you only publish “best Y alternatives” content with yourself on top, you are making a claim.
When you publish “best X alternatives” and consistently place Y alongside you, you are creating a relationship.
The model infers that relationship sideways.
So when it later encounters “best Y alternatives”, X does not feel new. It feels obvious.
This is why brute-force comparison content fails inside AI systems.
Self-promotion does not stick. Contextual adjacency does.
The fastest way to become the recommended alternative is not to attack the leader head-on. It is to first teach the model that you and the leader already belong together.
This week in AI search: 4 signals pointing to the same shift
GPT-5.2 rolls out
The biggest change is not intelligence. It’s decisiveness.
The model now commits to recommendations faster and with less hesitation.
If you are not already legible to the model, you are invisible at decision time.
Google’s Disco browser and AI expansion
Google is pushing AI from search results into the browsing layer itself.
AI is no longer answering queries. It is guiding sessions.
Discovery is being mediated earlier than ever.
Recipe creators seeing traffic collapse
These sites followed every SEO best practice.
AI still extracted the value and skipped the visit.
Owning the content no longer guarantees owning the user.
Stack Overflow’s AI Code Red shift
Stack Overflow is repositioning itself from destination to infrastructure.
Answers are no longer where users go. Answers come to them.
Even category-defining platforms are adapting to AI-first distribution.
The pattern behind all of this
All four stories point to the same reality.
Search was about retrieval.
AI is about resolution.
AI systems are not pointing people to the web. They are replacing parts of it.
When an AI resolves a question, it does not ask who ranks first.
It asks who it already understands well enough to trust.
The web is becoming training data. AI memory is becoming distribution.
And recommendations are the new rankings.
One question to sit with
If your top competitor disappeared tomorrow, would ChatGPT know where to place you without your website?
Most teams are still optimizing for clicks. A few are starting to optimize for memory.
The gap between the two is where the next category winners will come from.
More experiments soon! 👋