How to show up in AI search results

A prospective client asked me last month why their traffic was falling while their rankings had not moved. It is becoming the most common question I get, and the answer is that ranking and being read have come apart.

Someone types a question. An AI answer appears above everything else, written from several sources. The question is answered. The visit never happens.

You are still position three. Nobody is scrolling to position three.

What the goal is now

The old goal was to be the link somebody clicks. The new goal is to be the source the answer is built from, and ideally the name it mentions.

Those are not the same job. A page can rank well and never get quoted, because a model summarising four sources takes the sentence that states something clearly and skips the three paragraphs of preamble in front of it.

There is an uncomfortable implication for founder-led businesses. Being cited without being clicked still helps you, because your name appears in an answer at the moment somebody is deciding. But you cannot see it in your analytics, so it will look like your marketing is failing while it is working.

What gets a page quoted

Six things, in rough order of how much difference they make.

  1. Answer the question in the first two sentences. Not after your credentials, not after a paragraph about how important the topic is. If your answer is buried at paragraph six, a summariser will find someone whose answer is at paragraph one.
  2. Make claims that can be lifted whole. A sentence that survives being pulled out of your article and dropped into an answer is a quotable sentence. One that depends on the two paragraphs above it is not.
  3. Be specific. Numbers, named things, real constraints. “A team page can easily reach 76 MB if the headshots are PNGs” is quotable. “Image optimisation is important for site speed” is wallpaper.
  4. Say who is talking. An author with a name, a real biography, and a track record in the subject. Anonymous content is cheap to produce and is treated accordingly.
  5. Get mentioned elsewhere. Models learn what you are known for from the whole web, not just your site. Being described accurately in other people’s writing does more than another article on your own.
  6. Cover a subject properly. A site with twelve genuinely useful articles on one topic reads as an authority, which is why coverage beats volume. The same twelve spread across nine unrelated topics reads as nothing in particular.

You will notice none of that is a trick. This is the same advice that has been correct for a decade, except that filler now fails harder and faster than it used to.

What stopped working

Content written to occupy a keyword rather than to answer a question. When a model can generate a competent generic overview of any subject on demand, a competent generic overview has no value left in it.

The listicle with no opinion. “Ten tips for X” where the ten tips are the ten tips everybody lists. There is nothing in it to quote, because there is nothing in it that only you would say.

Volume as a strategy. Publishing more of what already fails to get quoted produces more content that fails to get quoted.

The advantage you have

Here is why I think this shift favours founder-led businesses rather than threatening them.

What cannot be generated is what you specifically have seen. The pattern you have noticed across forty client projects. The thing that went wrong and what it cost. The opinion you hold that half your industry disagrees with.

A large competitor with a content team producing thirty articles a month is producing thirty articles a model could have written. You are the only source for what happened in your business last year.

That is not a motivational line. It is a practical instruction about what to write: the specific over the general, every time, even when the general would get more searches. It also changes what an SEO budget is buying.

How to tell whether it is working

Measurement here is genuinely poor, and anyone selling you a confident dashboard is overstating what is available. Three things are worth doing.

  • Ask the assistants directly. Once a month, put your ten most important buying questions into the main AI tools and record who gets named. That is your ranking report now.
  • Watch impressions against clicks. Impressions holding steady while clicks fall usually means you are being summarised rather than losing position.
  • Ask new enquiries how they found you. The answer is increasingly “an AI mentioned you” or “I asked for recommendations”, and no analytics tool will tell you that.

Expect the reporting to be worse than it was. That is uncomfortable, and it is not a reason to go back to writing for a metric that no longer describes the outcome you want.

Where to start

Take the page that matters most commercially. Read the first two sentences.

If they do not answer the question the page exists to answer, rewrite them so they do. Then find every vague claim on the page and make it specific or delete it.

That is an afternoon, and it does more than a quarter of publishing. After that, one idea will carry you a month.


Not sure which part to fix first?

Most founder-led businesses are not short of marketing activity. They are short of a decision about what matters. The Founder-Led Marketing Scorecard takes about three minutes and tells you which of seven areas is holding the rest back.

    Leave a Reply