Clear entities, evidence and original information in AI discovery

Google says no special optimisation is needed for its AI features, while research suggests citations and statistics help. What is documented, observed and still a hypothesis.

When someone asks an AI assistant about your company or your field, the answer is assembled from sources. Marketing leaders want to know what makes a company a useful source.

There is no guaranteed formula. But three qualities come up repeatedly in official guidance and research: clear entities, visible evidence and original information.

This article separates what is documented, what has been observed in research, and what is still a working hypothesis.

What is documented

Google’s guidance on AI features in Search is the clearest official statement. It says the best practices for SEO remain relevant for AI Overviews and AI Mode, and that there are no additional requirements or special optimisations needed to appear in them.

To be eligible as a supporting link, a page must be indexed and eligible to appear in Google Search with a snippet. Google adds that meeting every requirement does not guarantee a page will be shown.

The same page lists practical steps. They include making important content available as text, making sure structured data matches the visible text on the page, and checking that Merchant Center and Business Profile information is up to date.

It also says you do not need new machine-readable files, AI text files or special schema.org markup to appear in these features.

Clear entities

An entity is a distinct thing a search system can recognise: your company, a product, a person, a place. Ambiguity makes recognition harder.

Google’s documentation on Organization structured data says adding this markup to your home page can help Google understand your organisation’s administrative details and disambiguate it in search results.

The markup can include your official name, logo, address, contact details and links to other profiles that describe you, through a property called sameAs. Google says there are no required properties and recommends adding the ones relevant to your organisation.

What this means in practice. Use one official company name consistently. Make sure your product names are not used for other things on your own site. Keep leadership names, locations and founding facts identical wherever you publish them.

This is a documented route for Google Search. Whether other AI systems use the same signals is not something most providers document in detail.

Visible evidence

Google’s people-first content guidance asks whether content provides original information, reporting, research or analysis, and whether it adds substantial value rather than rewriting other sources.

Research points in a similar direction. A 2023 paper on generative engine optimisation, by researchers including teams at Princeton University and IIT Delhi and presented at the KDD conference in 2024, tested ways of editing web content to increase its visibility in AI-generated answers.

The authors report that their top-performing methods were adding citations from credible sources, adding relevant quotations and adding statistics. These raised visibility by 30-40% on one of their measures. Traditional keyword stuffing often performed worse than the baseline.

This is observation, not guidance. The tests used a research benchmark and specific engines. Results varied by domain, and AI systems change frequently.

Original information

Original information is something a reader cannot get elsewhere: your survey, your benchmark, your project data, your expert’s first-hand explanation.

If many pages repeat the same facts, any one of them is replaceable. A page with a unique dataset or a clearly attributed expert view is harder to replace.

This is a hypothesis. It follows from Google’s guidance and from the research above, but no provider promises to cite original research.

What this looks like for a B2B firm

Hypothetical: a Japanese precision components maker wants to be cited when engineers ask AI assistants how to reduce vibration in robotic arms.

Entities. It standardises its company and product names across its website, distributor pages and trade directories, and adds Organization markup to its home page.

Evidence. It rewrites its technical guides to cite standards bodies and test results, with each statistic linked to its source.

Original information. It publishes anonymised results from vibration tests across real installations, with a named engineer explaining the method.

None of this guarantees a citation. It makes the company a clearer, more checkable source, which is useful for human readers too.

Who owns this work

AI discovery sits between several teams. SEO owns structured data, content owns evidence, communications owns how leaders and facts are described, and research or product teams hold the original data.

Name one person to coordinate. Their job is not to chase citations, but to make sure the company describes itself consistently and backs its claims.

The next article in this edition gives that person a practical starting point: an audit of whether your key company facts agree.

What to avoid

Promising inclusion. Any vendor or agency that guarantees AI citations is claiming more than any provider documents.

Special files and markup for AI alone. Google states these are not needed for its AI features. They may have uses elsewhere, but treat claims about them as unproven until a provider documents them.

Content that only rephrases others. Google’s guidance lists mainly summarising others without adding value as a warning sign.

Questions for your team

Before investing in AI visibility, ask three things.

Can a machine tell, from our own pages, exactly who we are and what we sell? Do our most important claims link to evidence a sceptic could check? What do we publish that nobody else could?

If the answers are weak, the first work is on your own site. That work also helps in traditional search, sales conversations and press coverage.

Keep watching

This area changes quickly. Google’s AI features page was last updated in December 2025, and other providers publish less detail.

Review your approach at least twice a year. Record what you change and when, so you can compare it with what you observe in AI answers over time.