Why your content isn’t showing up in AI search (and how to fix it)

A few months ago, I wrote about how AI is quietly rewriting the rules of search, and why that changes everything about how we build websites. You can read it here: “Building for the Visitors Who’ll Never Visit.”

That post was shaped by a website rebuild I was leading. Since then, I’ve spent more time in the data, and the takeaway is simple: this is not a niche concern for SEO people in dark rooms. It affects anyone who owns, runs, funds, or depends on a website.

Because the game is changing. Fast.

The numbers are stark

A growing share of searches now end without a click. People get an answer directly in the results page and move on. On mobile, it’s even more pronounced.

At the same time, AI-generated summaries are showing up more often, across Google and the newer “answer engines”. When those summaries appear, traditional organic click-through rates drop sharply, even for the top result.

And it’s not just Google. Tools like ChatGPT and Perplexity are now mainstream. Whether we like it or not, millions of people are using them as the first step for “find me an answer”, “explain this”, and “what should I do”.

The consequence is obvious: less traffic reaches websites, even when your content is good and even when you rank well.

The impact is real

We’re already seeing organisations report major drops in organic traffic. Some of that is normal algorithm churn, but the direction of travel is clear. If a user gets a complete answer without leaving the platform, a lot of websites simply never get the visit.

This isn’t limited to publishers. It hits e-commerce, professional services, education, B2B, charities, and public sector. If organic traffic plays any role in your growth, your recruitment, your sales pipeline, or your reputation, the ground is shifting under your feet.

Rankings aren’t enough anymore

Here’s the uncomfortable truth: being on page one is no longer the finish line.

In AI search, visibility is increasingly about being used , not just being found. Your content needs to be cited, referenced, and pulled into answers.

When AI systems generate responses, they draw from sources they trust. If you’re not one of those sources, you can be invisible, even if you technically “rank”.

That’s why the new question is not just “Do we rank?”, it’s “Do we get cited?”

What I learned along the way

As I dug into this, I kept finding things that surprised me. Things that are not in the standard SEO playbook.

1) Each platform plays by different rules

Optimising for one system does not mean you’re optimised for all. They draw on different source pools, weight signals differently, and surface content in different ways. A tactic that works for Google AI Overviews might do very little for Perplexity, and vice versa.

2) You might be blocking AI without knowing it

A lot of organisations added blocks for AI crawlers during the early wave of anxiety. Others have robots.txt rules that accidentally block more than intended. If AI systems cannot access your content, they cannot cite it. A five-minute check can be the difference between visible and invisible.

3) There’s a new file you’ve probably never heard of

It’s called llms.txt. Think of it as a way to point AI systems towards what matters most on your site. It’s simple, it’s lightweight, and hardly anyone is using it yet.

4) You don’t need to rank number one to get cited

One of the most interesting shifts is that citations do not always map neatly to the top 10 results. If your page answers the question clearly and is structured well, you can still be pulled into AI answers even if you’re not dominating rankings. That’s a real opportunity for smaller sites.

5) The “answer capsule” matters more than the whole page

AI systems look for quotable chunks, often something like 40 to 60 words, that directly answers the question. If the answer is buried, vague, or wrapped in hedging, you’re making it harder to be cited. The first 100 words now do an outsized amount of work.

6) Author pages actually matter now

Experience, expertise, and trust signals are no longer abstract guidelines. Clear author attribution, credentials, dates, and sources help systems decide whether your content is safe to quote. A proper author page is increasingly part of performance, not just good housekeeping.

7) Schema markup has a new purpose

Schema is not just about rich results. It also helps machines understand entities, relationships, and context. Organisation, Person, FAQPage, Article, HowTo, these are becoming a way to reduce ambiguity and increase trust.

8) Freshness is weighted differently per platform

Some systems strongly favour recency. Others blend authority and freshness. The result is that “what works” depends on where your audience is getting answers. A page updated last week might win in one place and make no difference in another.

What I’ve built

I’ve pulled all of this into something practical.

A system you can actually use:

A complete audit process
How to test visibility across ChatGPT, Perplexity, and Google AI Overviews using free methods. Build a query list, test systematically, establish a baseline, spot the gaps.
Technical setup guides
Copy-paste robots.txt configurations for AI crawlers, how to create an llms.txt file, schema templates for the types that matter, plus speed and rendering checks that still trip people up.
Content optimisation workflows
The answer-first structure that gets cited, how to rework existing pages without rewriting everything from scratch, question mapping templates, entity and semantic optimisation, and a sensible approach to keeping content fresh.
Authority building strategies
How to build the external mentions and references that act as trust signals, plus practical ways to strengthen author credibility and improve perception of expertise.
Platform-specific playbooks
Separate checklists for Google AI Overviews, ChatGPT, and Perplexity, because one-size-fits-all is not how this works anymore.
Monitoring systems
Weekly checks, monthly reporting, and quarterly reviews, with options for free and paid tools. Also, how to track progress when the metrics are no longer just “rank” and “traffic”.
Templates and prompts
Checklists you can duplicate, prompt templates to speed up the work, a quick reference card, and a plain-English glossary.

I wish this system had existed when I started. It would have saved me a lot of trial and error, and a lot of time spent joining dots that weren’t written down anywhere.

That’s why I’ve pulled everything I’ve learned into one place. It’s practical, it’s built for doing the work, and it’s the resource I kept wishing I could point to.

I’ll keep updating it as the platforms shift, new patterns emerge, and the tooling improves. And over time I’ll extend it too, adding more examples, more templates, and more platform-specific guidance as we learn what actually moves the needle.

Here is a link to the Notion guide, let me know how you get on.