What AI says about your brand’s weaknesses and how to check this

Below, we explain how to find out what AI is saying about your brand, how to distinguish between a criticism and a mistake, and what to do about each.

Where do AI chatbots get their criticisms of brands from?

From the same sources as the rest of its answers: from knowledge gained during training and from web pages retrieved at the time of the response. When the model uses a search, it breaks the question down into several sub-queries, including those relating to price, specifications, reviews and availability. The query regarding reviews leads it to reviews, forums and comparisons, and from there to criticisms. As the model formulates its answer, it describes the text How AI chatbots choose the brands they recommend.

This is also evident in the search engine data. In Google Search Console, from July to October 2026, queries appeared for Aelo’s pages that all followed the same pattern: ‘Brand X is being criticised for Y, which is putting some customers off – what are the reasons for this? The ‘Y’ slot was filled with, amongst other things, infrequent updates, a rigid process, an unclear scope of services, geographical restrictions, a difficult start-up phase and opaque pricing. People don’t usually type in queries like this. They look like queries from AI tools that check allegations against brands before writing about them.

How can you find out what AI says about your brand’s weaknesses?

Ask the AI chatbots the questions that a customer comparing offers would ask, and record each claim along with its source, if the model provides one. Ask the questions across several chatbots, with each question in a new conversation and without suggesting any answers. The general testing method is described on the website How to find out what AI says about your business. If the chatbot doesn’t recognise the brand at all, asking about faults won’t help: start with the page ChatGPT doesn’t know my company.

Record not only the allegation itself, but also whether the model cites a source and how often the allegation recurs. A single mention in a single response may be a coincidence, as chatbots respond slightly differently each time. An allegation that recurs in many responses and across several models is based on a consistent source. The text explains why a single response proves nothing. Why does AI give a different answer every time?.

An allegation, outdated information or an error: how can you tell the difference?

Next, whether the claim is true today. A true claim describes something that customers actually report. An ‘out-of-date’ claim was once true, but the company has since rectified it. A factual error is information that was never true, such as a mistaken industry classification or a competitor’s price. Errors are dealt with in a separate article: AI is providing false information about the company. Here, we focus on opinions and allegations.

What should one do about an accusation that is true?

Respond to it with specific details in places that potential clients read, rather than glossing over it. If clients complain about a lengthy start-up phase, describe on the website how long the implementation takes, what it involves, and what can speed it up. If the complaint concerns the price, show what the offer includes and who it is aimed at. Models quote statements that can be cited: a figure, a deadline, a scope. The absence of such facts on the website is what we call content gap. A good Google ranking isn’t enough, because the model also draws on information from other brand-related websites; more on this in the text Visibility in AI and SEO.

The second step involves customer reviews on independent websites. A consumer who sees one complaint and a dozen or so specific, positive reviews is less likely to repeat that complaint as the brand’s defining characteristic. This is the same process that leads to recommendations, as described in the text How to get AI to recommend your company.

What should be done about a charge that is no longer relevant?

Find the source and update it wherever possible. Ask the AI chatbots for the source of the criticism, and if the model uses a search function, check which websites it cites when asked about shortcomings. Often, this is an old review, a comparison from before a change to the offering, or a forum thread. Where possible, add up-to-date information or ask the author to update it. On your own website, describe the change with a date, e.g. „since March 2026, integration with X has been available on every plan”, as the date helps the model recognise the information as more recent.

The effect won’t be immediate. Changes to the sources that the model searches on the web take effect more quickly than changes to the knowledge gained from training, but they still take weeks. This is explained on the website Why do changes take time?.

How can you measure what AI says about your brand over time?

The same set of questions at regular intervals, with an assessment sentiment for each response. Sentiment indicates whether the model describes the brand positively, neutrally or negatively, whilst the list of complaints shows which of them disappear following your actions. The set of questions must remain consistent, as this is the only way to ensure that subsequent measurements can be compared. This principle is described by the term frozen query set, and the frequency per page How often should visibility be measured?. How we calculate sentiment and other indicators is described in methodology.

The criticisms also influence whether a model will recommend a brand. A brand that a model is familiar with but describes with reservations is more likely to end up in the situation described in the text Knows the brand, but doesn’t recommend it, and its place is taken by a competitor. The text explains why this happens Why does AI recommend the competition?. Agencies may include sentiment data in their monthly reports, as we discuss in the text AI Visibility Report for the client.

How does Aelo help you find out what AI is saying about your brand?

Checking what AI says about my brand only makes sense as part of a series of measurements. We measure whether ChatGPT, Claude, Gemini and Perplexity mention your brand, how they describe it and which brands they recommend instead, and the report shows mention, citation, sentiment and sources. The measurements are carried out by the module Visibility, the tasks are sent to Tickets, and the materials addressing the allegations are prepared by Studio. The text describes the whole process How Aelo works.

You’ll carry out the first measurement for a single brand on Free account. Agencies can find out more on the website Aelo for agencies, and local businesses on the website Aelo for small businesses.

Try it for free

FAQ

Sources

• AirOps, report on searches, follow-up queries and citations in ChatGPT (March 2026): https://www.airops.com/report/influence-of-retrieval-fanout-and-google-serps-in-chatgpt
• Google Search Console for aeloapp.io, search queries from 20 July to 16 October 2026 (own data, excluding brand names)