Louis Ashworth Published September 23 2026 There’s not much inherently wrong with using Chat GPT as a search tool. In terms of getting a quick answer to something, it’s functionally similar to Google, now that Google has decided the best approach is to précis all search results with a big box that frequently and confidently lies to you. Either way, you’ll get a confident answer you can’t trust in a couple of seconds rather than relying on your own judgment to find one you can possibly trust in several seconds.

We’re only human. We do what we can. There are certain contexts, however, in which you might want to disguise the fact that you used a word-probability engine to get an answer for something.

For instance, while making a financial filing. Despite that, plenty of people are happy to flaunt their usage of the tool: Some content could not load. Check your internet connection or browser settings.

We should be clear: these citations are not coming from top-tier companies. A flick through the list throws up non-household names like China Tontine Wines Group, IM Cannabis Corp, Coolbit Technologies and THG. There’s a noticeable density of Indian and Israeli companies.

Chat GPT usage tends to occur in offhand references to unserious topics like total addressable markets, and usually in the more narrative — and, therefore, less risky — parts of filings, where companies talk about wishy-washy things like developments in key markets and further growth prospects. Maybe these sections are increasingly written using AI, maybe they’re increasingly written by AI. Either way, they’re a slop-friendly vestibule.

In most examples we found, there was just one “utm_source=chatgpt.com” in any given filing. Sometimes, such as in this 6-K from miner Critical Minerals Corp, usage appears extensive: Sometimes, such as in the 2024 annual report of Octopus Renewables Infrastructure Trust, Chat GPT usage seems to be the least of the issues (this link is definitely supposed to go to the Financial Times, at ft.com, rather than STMicroelectronics, at st.com): Sometimes, such as in this annual report by Sports Hero, the degrees of separation from an actual source look increasingly stretched: A study sent to FT Alphaville by Originality.ai, an AI plagiarism detector, found mentions of “utm_source=chatgpt.com” across 44 SEC documents, noting a greater preponderance of other AI writing tells in those relative to a random sample. Originality noted the example of Bio Key (Cayman), a pharma company spun out of ABVC Biopharma.

In a June filing, Bio Key said it was “positioned in a global nutraceuticals market projected to grow from ~$112B in 2022 to ~$167B by 2030” — citing links to Grand View Research and Allied Market Research, the latter of which carried the Chat GPT-slug-addendum calling card: Neither link supports the “global nutraceuticals market” figure mentioned. The former (Grand View) claims it will be $919.1bn by 2030, which the second (Allied Market) suggests a much bigger addressable market of $413bn currently, and $650.5bn in 2020: We’re sure Bio Key would have loved to indicate a bigger industry size, but hallucinations might cut both ways. Later in the same filing, Bio Key correctly cites the Grand View numbers.

We asked the company what happened; its people had not responded at time of pixel. As we must so often say, none of this really matters. Filing bumpf was bad before LLM proliferation.

It might be worse now, it might be better — either way, it’s not something investors ought to pay much attention to. In fact, we’ve often wondered why companies even bother. It looks like some probably don’t.