Reputation · AI answers

Can you influence what AI says about your brand? Errors, risks and the five things that work

By Kristian Stig Henriksen · Dear Future

Yes, in part. What an AI assistant says about your brand comes from two sources: what the model learned during training, and what it finds when it searches the web while answering. The first you cannot correct. The second you can influence, because it is built from pages that exist right now, and most of them are your own or written by someone who mentions you.

Peach and orange smoke billows across a grey-blue background, soft and without hard edges.

Where the answers come from

A language model is trained on enormous amounts of text up to a fixed date. What it knows about your brand from training is a compressed average of everything written about you when it was trained, including the outdated and the wrong. With web search switched on, the assistant additionally fetches current pages and writes the answer from them with references. Those are the answers customers get when they ask about something concrete, and those are the ones that can be moved.

So the question is not whether AI can be influenced, but which part. The model's memory: no, not by you, and only slowly by the world. The model's retrieval: yes, at the same pace at which you can change the pages it finds.

Where errors arise

Pure inventions, hallucinations, do occur, but they are rarer in answers with search, and they are harder to prevent than errors with a source. Most wrong answers about brands have a source, and the source can be found.

The five things that work

  1. Own the basic facts. One page that says clearly what you do, for whom, where and since when, with a date. Structured data for name, address, logo and profiles.
  2. Answer the questions customers ask. Not about you, but about the need: what does it typically cost, what is the difference, who suits what. One page per question, with a short, direct answer at the top.
  3. Consistency across sources. The same description on your site, LinkedIn, directories and at partners. Models weight agreement between sources.
  4. Mentions at the sources the models already cite. Find them in your category and get named there: industry overviews, trade media, partners. It is not link building; it is being present where the answer is fetched from.
  5. Measurement. Put your customers' questions to the assistants regularly and read what they answer before your customers do. Without measurement you discover the error when a customer mentions it.

What you cannot do

You cannot call someone and have an answer corrected. You cannot remove what the model learned in training, and you cannot ask it to recommend you. Nor can you control the tone of a single answer; the same question tomorrow gives a different answer. What you can do is change the ground the answers are built on, and measure whether the average moves. That is work over weeks and months, not days.

Is it a fad, or is it here to stay?

Adoption says it is here to stay. Denmark sits at the top of the EU in the use of generative AI, and the assistants are used to research purchases, not just to write text. Gartner predicted as early as 2024 a 25% drop in classic search volume by 2026 because of AI assistants. Whether the numbers land exactly there matters less than the direction: a growing share of buyers' questions is answered in one answer with a few names. That is the reason the question in the headline is asked at all.

Read on: get recommended by ChatGPT, what an LLM is and how to measure AI visibility.

From worry to numbers

What are the assistants saying about you right now?

Get an AI visibility analysis: we put your customers' questions to the assistants every day and show who is named, cited and recommended, and in what tone.

Get the analysis

Frequently asked questions

Can I actively influence the answers AI assistants give about my company?

In part. You cannot change what the model learned in training, but you can influence the answers built on web search: by owning the basic facts on your own pages, answering customers' questions, keeping descriptions consistent across sources, and being mentioned at the sources the models cite.

What can go wrong if I do not monitor what AI says about my brand?

Outdated or wrong information can be repeated to customers without you knowing, and competitors can be recommended on questions where you would be the obvious answer. Because the answers are not visible like a results page, it is typically discovered only when a customer mentions it.

How do I protect my brand against wrong information from AI?

Find the source of the error; most wrong answers have one. Correct your own pages, date them, ask third parties to update theirs, and publish a clear, dated answer to the question the error appears in. Then measure whether the answers change over the following weeks.

Is AI visibility a lasting strategy or a passing fad?

The use of AI assistants for purchase research is growing, and Denmark sits at the top of the EU. The direction is clear even as the numbers change: a growing share of buyers' questions is answered in one answer with a few names. Measuring whether you are among them is therefore a lasting discipline, alongside measuring your search visibility.

Sources

  1. 1.Google Search Central. AI features and your website. Google Search documentation.
  2. 2.OpenAI (2024). Introducing ChatGPT search.
  3. 3.Gartner (2024). Gartner Predicts Search Engine Volume Will Drop 25% by 2026, Due to AI Chatbots and Other Virtual Agents. Press release.
  4. 4.Eurostat. Digital economy and society statistics.

Back to the front page