AI Chatbots in Valuations: What Collectors Need to Know

“But ChatGPT told me my vase was worth £100,000!”

Sound familiar? As a professional valuer, the most common question I receive from clients after submitting a valuation is why my analysis differs from what they found on an AI chatbot.

With the rise of artificial intelligence (AI) over the past five years, the landscape for technology as we know it is rapidly evolving, affecting every industry, including valuations. These days, it is very difficult to escape AI; many use it without even realizing (even a simple Google search uses AI to produce the Gemini summary that appears at the top of the results page). The art market has picked up on this, as is reflected in the recent surge of art tech startups that have emerged post-pandemic. In the valuations field, there are three primary areas heavily incorporating AI: databases for auction results (Artnet, Artprice, MutualArt, etc.), reverse image searches, and, lastly, the (in)famous(?) AI chatbots. Each category deserves its own article, but AI chatbots tend to cause the most confusion, so they will be my focus here. There is currently very little literature on the use of AI chatbots for personal property valuations, especially from the perspective of a valuer.

Why Are AI Chatbots Important?

It is crucial to first briefly outline how AI chatbots operate and what makes them so influential. AI chatbots are having an astronomical impact in reshaping the way users approach research. Nearly 30 years ago, Google and other search engines were revolutionary in collecting and presenting information in an organized format. Fast forward several decades, AI chatbots now go a step further to synthesize collected information by reading its contents for us and summarizing them so that users aren’t required to reference the source material at all. For simple tasks, this can suffice, but for more complex work such as valuations, confusion and errors often ensue when there is no human input to verify AI’s work.

Why do AI Chatbots Make Mistakes?

There is a misconception that because a computer generated an answer, the answer is objective and consistent, but AI chatbots are the first to remind users of accuracy disclaimers. The reality is that chatbots are as accurate as the people who programmed them (these people are not art valuers!), and much like individuals, no two chatbots are the same. This means that when asking multiple chatbots the same question, they will often produce entirely different answers, ranging widely in accuracy. Chatbots are programmed to have different strengths and weaknesses depending on what their programmers prioritized, and some will not engage with valuations at all, (quite wisely) redirecting the user to professional valuers. As a professional valuer, I would even be wary to claim that there is one AI chatbot best for art historical research because the technology is shifting so swiftly that within a week of publishing this article, a recommendation may already be out of date!

When clients ask me about AI chatbots, here is my advice:

Why AI Chatbots Can’t Replace Valuers:

Basic object identification:

For very common household items or famous works of art (think Van Gogh or Picasso), AI chatbots can often correctly identify photographs. However, the moment the artist or subject matter becomes any more obscure, roadblocks begin to appear. As an Asian art specialist, I frequently find that clients misidentify their objects because a chatbot couldn’t correctly identify the materials, country of origin, religious subjects, etc. These tend to be mistakes that human specialists (who identify objects for a living) can remedy immediately.

Dating:

By far the most common issue I encounter with clients is that an AI chatbot told them their object was an antique, when it is in fact a Modern or Contemporary reproduction. The main reason for this error is that chatbots heavily rely on textual information for dating objects, so visual analysis receives little (if any) weight. Some chatbots now provide elementary visual descriptions, but their detail and accuracy are inconsistent and fall short, whereas professional valuers are expected to do both. For Asian antiques and other specialties, this type of error can be detrimental in determining a value.

Comparables:

As mentioned above, AI chatbots are not advanced enough yet to accurately visually analyze a single item, let alone visually compare multiple items, which is a crucial step for selecting appropriate comparables. Chatbots usually rely on text alone, which is insufficient for comp selection; a set of human eyes is required to make this call. AI chatbots may be able to locate links to sales of objects with the same text description, but these often look nothing alike. This type of mistake can be disastrous when chatbots rely on sales data from the wrong comparables to determine the value of an item. Additionally, AI chatbots exclude offline comparables that professional valuers know how to find, which can be crucial in valuations. Mistakes tend to be even more severe for specialized objects.

Market analysis:

Market analysis can be accurate for Contemporary blue-chip artists or other household names, but only for their general market trends, with no further nuance. As expected, for many specialist areas, the technology is not yet robust enough to provide accurate market analysis conclusions. This is due to the lack of visual analysis, chatbots not searching for sources in the correct places, or the inability to analyze results. Collectors still need a professional valuer or advisor to make accurate judgments.

Values:

The above elements all contribute to the value conclusion of an item, and if these steps of the equation are incorrect (which they often are!), there is little chance of accuracy in determining values. Even when these aspects are correct, AI chatbots are not equipped to weigh the nuance of each component.

Not knowing the right questions:

The answers from AI chatbots are only as good as the questions they are given. Collectors often (understandably) do not know which questions to ask, resulting in less helpful AI output. Professional valuers do not need AI chatbots, but if we did choose to consult them, we would produce more helpful information because we have the academic specialism and the professional experience to anticipate the types of queries associated with each type of object. Often, clients receive the wrong information because they do not have the baseline knowledge to use AI chatbots as an effective research tool.

It’s Not All Bad: Pros of Using AI Chatbots:

Object identification for very common objects:

As previously mentioned, very common objects with a large internet presence are usually retrievable on AI chatbots.

Historical information:

For casual research, AI chatbots can be a great way for collectors to learn more about the history of their items.

Signature recognition (sometimes!):

AI chatbots can sometimes help decipher sloppy handwritten signatures. They are not, however, authentication tools, and the AI suggestions given are merely possibilities for what the signatures might say.

Language translation of clearly printed materials:

For modern, clearly printed texts, AI chatbots can sometimes identify the language used or give the user an overall gist of a text, akin to using Google Translate. This usefulness evaporates immediately when applied to handwritten or antiquated texts. As an Asian art specialist, I frequently encounter clients mistranslating texts found on antiques, especially with calligraphy. A professional valuer and/or translator is always necessary to ensure accuracy.

Will AI Chatbots Replace Professional Valuations?

Not anytime soon, and it is currently difficult to imagine a scenario in which AI’s capabilities outperform professional expertise, especially in specialized areas. AI is a tool, not a decision maker. Much like Google, AI chatbots are a fantastic resource for collectors to use for casual interest, but, due to their limited research capacity, frequent errors, and lack of nuance, chatbots are not suitable for valuations and should never override a professional opinion. AI also cannot be held accountable for its errors, whereas professional valuers undergo rigorous training and adhere to strict ethical guidelines that ensure accuracy and responsibility. As a result, insurers, governments on both sides of the pond, and arts institutions reject AI-generated valuations in favor of professional ones, and with good cause.

Although my initials are AI, rest assured my valuations are anything but artificial. If you are searching for a professional valuation, Doerr Dallas is here to help!

Disclaimer: This article reflects current technology and valuation practices as of September 2025 and may not apply in all jurisdictions.