How to Get AI to Recommend You: Content & Technical SEO

Being found is not the same as being cited. Being cited is not the same as being recommended.

In our previous Mediaworks Masterclass, we looked at how Digital PR builds the third-party evidence that AI platforms rely on when deciding who to recommend. The response told us something useful: when we asked attendees how confident they were that ChatGPT, Gemini or Copilot would recommend their business, half weren't confident at all. Only a handful said they were.

We asked the same audience what they thought was actually stopping AI from recommending their business. The two biggest answers were off-site mentions and authority, and website and content. A third answer came up almost as often: "we don't know."

That third answer is the one worth addressing first, because it points to a gap most businesses have. Few have properly audited whether their content and technical setup are even capable of being read, understood and reused by an AI system. This post covers that ground: what decides whether you get cited at all, and the content and technical fundamentals that make it possible.

The three things that decide an AI citation

Before any content or technical fix, it helps to know what's actually being evaluated. There are three factors.

Identity is about who you are in the market and how you're different. Is your price the reason to choose you, your lead time, your customer relationships? This is the strategic groundwork that shapes everything else, because a model has no reason to recommend you over a competitor unless something distinguishes you.

Extractability is whether your content and site can actually be read and lifted by an AI system. Is it written sensibly, structured to be scanned, marked up so a model can find the answer? This is where content and technical work meet.

Corroboration is whether other credible sources describe you the way you describe yourself. If your brand claims to be reliable or sustainable, are there third parties confirming that? This is the Digital PR pillar we covered previously.

Get identity and corroboration right but fail on extractability, and none of it reaches the model. That's why crawlability and indexation come first.

Crawlability and indexation: the entry ticket

If a crawler can't reach your page, nothing else matters. Not the quality of the content, not the links pointing at it. This is the fundamental layer, and it's also the one businesses most often get wrong without realising it.

Four checkpoints are worth auditing:

• Crawler access. AI crawlers such as GPTBot need explicit allow rules in your robots.txt. If your robots.txt has no disallow rules at all, you're likely fine. If it does, add specific allow rules for the bots you want to reach you. This is a two-minute job with no development cost.

• Rendering. If your site relies on client-side JavaScript with no server-side or pre-rendered HTML, an AI bot may not be able to read it at all. A site can look excellent to a human visitor and still be functionally invisible to an LLM. These are two different audiences with two different requirements, and a good user experience doesn't guarantee a readable one.

• Sitemaps and structure. Your sitemap is one of the first places an LLM goes to understand how authority is distributed across your site, so it needs to be accurate and current. A logical URL structure matters here too. LLMs are resource-intensive to run, so they favour sites that are easy to navigate over ones that require extensive crawling to make sense of. A well-organised site is cheaper for a model to process, and that can be the difference that gets you prioritised.

• Indexation hygiene. Incorrect canonical tags, orphaned pages and misapplied no-index tags all quietly block content that should be visible. This is worth checking before any investment in content or outreach, because none of that activity improves visibility if the technical foundation isn't in place.

Get this wrong, and you're not in a weaker position in the same game. You're not in the game at all.

Structuring content for extraction

Once a page can be crawled, the next question is whether it can be lifted cleanly into an answer. This isn't about tone of voice or writing style; it's about structure.

Answer the question early. If a paragraph is going to answer "how do I pick a destination for my next holiday," the answer should be in the first sentence, not buried in the third. A model wants a complete, quotable answer, and it wants it fast.

One clear aim per page. This doesn't mean every page has to be short. A guide to choosing a washing machine should cover cost, sustainability and household size, but everything on that page should build toward answering the one overarching question the page exists to answer.

Self-contained answer blocks. FAQ-style content that answers a question fully within itself, without assuming the reader has seen another part of your site, is the easiest format for a model to lift into a short prompt response or an AI overview.

Clear heading hierarchy. One H1, one core question. H2s to break that question into logical parts. H3s to add detail. This sounds basic, and it is, but it's also one of the most commonly overlooked fundamentals. A significant share of AI-generated answers are pulled from the top third of a page, so don't bury the answer under a build-up.

A useful test: write for extraction, not just for reading. Context and nuance can still follow underneath, but the answer needs to come first.

Schema, entities and technical health

This is where content and technical work blend, and where you tell a model, in language it can act on, who you are.

Structured data. Schema markup is low-effort and low-resource to implement, and it gives an LLM a clear, machine-readable way to extract information directly from your site. There's ongoing debate about how much weight schema carries, but the cost of adding it is low enough that it remains worth doing.

Entity consistency. This is about building recognition not just for your brand, but for the people and assets behind it. If you have spokespeople, sustainability leads or product specialists, make sure they're identifiable on your site as the authors of content or the source of quotes and statistics. The question worth asking is: what do you want to be famous for? Many organisations have a clear tone of voice document but haven't audited whether every part of the business is actually reinforcing the same message. Consistency and repetition across every touchpoint, on-site, in social, in third-party coverage, is what builds a recognisable entity a model can point to.

Core technical health. Page speed, Core Web Vitals, mobile rendering and canonicalisation issues all need to be clean. These aren't new SEO concerns, but they remain a precondition for AI visibility.

Internal linking. If your core commercial pages are consistently linked to, that signals where your authority sits and helps a model understand the natural next step to recommend. Think about how a ChatGPT conversation typically progresses: someone states a problem, gets pointed toward possible solutions, then wants to know what to do next. Clear internal linking is what lets a model carry that authority through to the conversion point.

Depth, freshness and consistency

Answer-first content doesn't mean shallow content. Following the May 2026 core update, longer, more thorough pages have been prioritised. The two ideas aren't in tension: answer the question early, then use the rest of the page to add genuine depth, linking out to more specific pages where needed.

Freshness matters because this isn't a set-and-forget exercise. Models and their sources shift constantly, so returning to your key content and updating it needs to be a regular part of the plan, not a one-off project.

Consistency ties back to the same question raised earlier: what do you want to be known for? The more consistently that message is repeated across your site, your spokespeople and your third-party coverage, the easier it becomes for a model to build a reliable picture of your brand.

Key takeaways

• Crawlability and indexation determine everything else, no matter how strong the content is

• Structure new and existing content so it can be lifted cleanly into an AI answer

• Use schema and structured data to make your claims machine-readable

• Check for the technical issues most likely to be quietly blocking citation

• Run a first-pass content and technical audit against your own site using these fundamentals as the checklist

Want to know where you stand?

Request a FREE AI Visibility Audit from Mediaworks.

Using the Authority pillar of our MX framework, we'll test your brand against real prompts, assess how you compare to your competitors, and identify the priority actions that could improve your chances of being recommended.

Request your FREE AI Visibility Audit