The test that predicts whether AI will cite your page

Before a single buyer prompt is run, you can predict fairly well whether an AI engine will cite a given page. Three things decide most of it: whether the page can be read at all, whether it is structured for machine extraction, and whether your brand carries authority on the topic.

Score well on all three and you show up. Score poorly on any one and you are filtered out before the model even weighs your content.

Here is the framework, the evidence behind each part, and how to run the test on your own pages today.

The three inputs that predict citation

Readability by machines. If a model cannot render your page cleanly, it cannot cite it. Technical barriers like blocked rendering, heavy client-side content, and confusing canonicals are among the most common reasons a page fails to get cited. Server-side rendering and fast load are table stakes.

Structure for extraction. Models cite what they can lift in a clean unit. Structural clarity, clear headings, and factual specificity with verifiable data all increase the odds of being cited, and schema markup is what makes those signals explicit.

Topical authority. This is the heaviest of the three. Topical authority is the strongest single predictor of AI citation, and only 12% of AI-cited links rank in Google's top 10, which means classic SEO rank is a weak proxy for citation. Being a recognised entity on your topic matters more than ranking for a keyword.

Run the test on your own page

Give each page a quick score out of 100, weighted toward authority since the evidence says it matters most.

• Render and speed, 25 points: does the page return full content without JavaScript, and does it load fast. Open it with scripts disabled and see what is actually there.
• Structure and schema, 25 points: one clear H1, scannable headings phrased as real questions, a direct answer near the top, and valid Article, Organization, and FAQ schema.
• Topical authority, 50 points: does your brand have depth on this topic across multiple pages and third-party sources, and is it a clean, disambiguated entity that a model can recognise. Thin, one-off pages on a topic you have no presence in will score low here no matter how well built they are.

A page that scores high across all three is citation-ready. A page that aces structure but has no topical authority behind it will still struggle, which is why owned content alone is rarely enough.

Where readiness ends and earning begins

The test predicts whether a page can be cited. It does not place that page into the competitive answers that drive deals, because those answers are built from third-party editorial and review sources. Readiness gets you eligible. Getting cited on your money prompts still takes appearing in the specific sources the AI pulls from for them.

Haystack runs that second half. It measures where you stand on each buyer prompt, names the exact sources the AI cites to recommend competitors, drafts the pitches that earn you a place in them, and proves when a placement turns into a new citation. The readiness test makes your pages eligible. Haystack closes the prompts.