Authority. Get the world to confirm your brand.

Engineer · Authority

AI assistants build their answers from what other sources say about you. Authority building gets the sources they trust to name you, describe you correctly and recommend you.

What is authority building for AI search?

Authority building for AI search is the work of earning mentions, citations and confirmed facts about a brand on the third-party sources that AI assistants read and cite. It covers digital PR, mention building, communities, reviews, owned social channels and the entity layer of Wikidata and structured data.

It is the third work area of Generative Engine Optimization, next to Technical and Content. Technical makes your brand readable. Content makes it quotable. Authority gets the rest of the internet to confirm it.

Your website is the smallest lever

An AI assistant does not rank your pages. It reads what the internet says about you and then it judges.

  • 85% of brand mentions in AI answers trace back to third-party content. The brand's own domain accounts for 13% (AirOps research, presented at BrightonSEO, October 2026).
  • In our own panels, the brand's own site gets a median of 3.4% of all citations on non-branded prompts. That is across twelve client and prospect panels and 325,000 citations. In nine of the twelve, the share is 5% or lower.
  • Lists, comparisons and reviews are about one in ten cited pages, and they sit at the top. In six of seven panels, the most-cited list gets at least 70% of the citations of the most-cited page of any kind.
  • On one client panel, the brand was named in 96% of answers to brand prompts and in 17% of answers to category prompts. The model knew the name. It did not use the brand to answer the category. We showed this case in our second webinar.

The gap between 96% and 17% is an authority gap. The model has not seen enough independent sources connect the brand to the category.

Start with the citation source list

We do not start from domain authority scores. We start from the answers.

  1. Sample. We run your prompt panel across ChatGPT, Gemini, Google AI Mode, AI Overviews, Claude and Copilot, many times.
  2. Extract. We collect every cited URL and classify it: press, comparison site, review platform, community, video, directory, competitor.
  3. Rank. We rank the sources by how often the engines cite them in your category, per engine and per market.
  4. Compare. We mark where competitors are named and you are not.

That list is your authority plan. It tells you whether your category runs on trade press, on review platforms, on Reddit threads or on YouTube. Every category is different, and every engine is different.

Six channels, one entity

  • Digital PR. Editorial coverage in the publications the engines cite, with a named expert in every story.
  • Mention building and listicles. Your brand in the comparison articles and "best of" lists that answers are built from.
  • Community management. Named experts taking part in Reddit, forums and publishing platforms, openly.
  • Review management. Reviews that are independent, recent and readable by machines.
  • Owned social. YouTube, LinkedIn and other channels run as sources a machine can read.
  • Wikidata and Wikipedia. The identifier every other profile points to.

You rarely need all six at full strength. The citation source list decides which two or three carry your category.

The Entity Map holds it together

Six channels can describe six different companies. A journalist uses last year's boilerplate. A review profile lists an old address. A LinkedIn page says something else again. Each mismatch teaches the model a slightly different brand.

The Entity Map is the single record of who you are to a machine. It holds every entity you want recognised, the facts that describe it and every place those facts must match. We publish it on your site as structured data and connect it to your profiles with sameAs links.

Organisation first, then authors

Machines trust organisations through people. A named expert with an author page, a consistent profile and a record of publications gives every claim a source.

The hierarchy runs from the organisation to its brands and products, and to the authors who speak for it. We build that hierarchy in structured data. Then we help the people in it become authors the press and the engines recognise. That programme is author enablement.

What we do not do

  • No fake or bought reviews. They are banned under EU consumer law and they poison the source the model trusts most.
  • No undisclosed accounts. Every community post comes from a real person or an official account, with the employer stated.
  • No self-ranked "best of" pages on your own site. They get cited but rarely earn the recommendation. We place you on third-party lists.
  • No edits to your own Wikipedia article. We propose changes in the open, with the paid relationship disclosed.
  • No guaranteed mentions. AI answers are probabilistic. We report rates across many runs.

How we measure it

Leading indicators show the work is landing: placements in target sources, new referring mentions, review volume and recency, answered threads.

Lagging indicators show it moved the answer: mention rate on category prompts, citation rate, Share of Model, sentiment and accuracy. We sample the same panel on the same engines before and after. We report prompts answered from retrieval separately from prompts answered from the model's memory, because they move at different speeds.

If a channel does not move a metric in two quarters, we stop it.

How it runs

Authority work is part of the Engineer package, from €3,000 per month. Your team, our team, your PR agency or a partner we bring in can execute it. Media and placement costs are budgeted separately. You see every cost before we commit to it.

The rhythm is weekly monitoring and reactive work, a monthly report, and a quarterly re-baseline of the prompt panel.

How is authority building different from link building?

Link building counts links and domain scores. Authority building for AI search counts mentions in context on the sources the engines cite. In an Ahrefs study of 75,000 brands, brand mentions on the web correlated much more strongly with visibility in AI Overviews (0.664) than backlinks did (0.218). A link still helps. A mention without a link also counts.

How long does authority building take?

It depends on how the prompt is answered. If the engine retrieves live sources, a new placement on a cited domain can change the answer within weeks. If the model answers from memory, change takes months and usually a model update. Our baseline classifies each prompt, so you know which case applies.

Do we need all six channels?

No. The citation source list shows which sources carry your category. A B2B software company usually needs trade press, review platforms and LinkedIn. A retailer usually needs reviews, comparison sites and video. We start with the two or three channels with the largest gap.

Can you guarantee that ChatGPT mentions our brand?

No, and nobody can. AI answers vary from run to run. We measure how often you are mentioned and cited across repeated runs, change the sources that drive that rate, and show the movement.

We already have a PR agency. Does this replace it?

No. Your PR agency keeps the relationships and the campaigns. We add the citation source list, the Entity Map and the measurement. Most clients have us brief the agency on which outlets and which wording move AI answers.

See which sources carry your category

The GEO baseline shows which pages the engines cite for your prompts, and where your brand is missing.