Hey folks,
Elliot here.
What do you do when LLMs start giving prospects incorrect information about your brand?
Maybe you’ve changed your product, but ChatGPT and Gemini are still surfacing outdated information. Or, worse, a competitor has published something about you that simply isn’t true and it’s now appearing on LLMs.
That last one happened to one of our clients recently.
A competitor had published an article containing objectively incorrect information about them. This was beyond throwing subtle shade…it included claims that were factually wrong and potentially very damaging.
The problem came when those claims started appearing in LLM responses and AI Overviews.
In this newsletter, I’m going to show you how we solved this misinformation problem within 24 hours, and how you can use the same approach to make sure LLMs have an accurate view of your brand.
The problem: a competitor making damaging claims about our client
For context, our client works in payments, an increasingly competitive and fast-growing space where claims about their capabilities can have a real impact on whether a prospect considers them.
I’ll keep the exact claims and brands anonymous, so let’s say the main claim was something like this:
“[Our client] is not licensed to offer services in North American markets.”
This was objectively false. The article also made some softer claims that were still damaging. It suggested our client had “limited geographic reach” when they operate globally, alongside vague claims that the product itself was basic.
We started seeing LLMs cite the article when answering questions about our client.
The same misinformation was also appearing in AI Overviews for bottom of funnel searches like “[Our client] vs [Competitor]”, exactly the kinds of searches prospects make when they’re actively comparing solutions.
Part of the problem was that our client didn’t have a clear answer to these questions on their own website. And to be fair, they’d never really needed one before. These were simply facts about the business that didn’t feel necessary to spell out explicitly.
But in the absence of a strong first-party source, LLMs were working with whatever information they could find. In this case, the clearest source available was the competitor’s article.
What we did about it: creating a GPT article to set the record straight
The good news is that LLMs and AI search experiences can surface newly published information fast: typically much faster than SEO.
So our job was to simply fill that gap with a better source, which is exactly what we did with a GPT article.
GPT articles are our proprietary framework for building LLM visibility. The basic idea is to create short, highly specific pieces of content that directly answer the kinds of questions your buyers are asking LLMs.
For our purposes here, we didn’t need to create some enormous guide covering their licensing, geographic coverage and every product capability under the sun, which probably would’ve taken weeks—time in which they could be losing customers.
Instead, we created a simple page designed to answer the exact question where the misinformation was appearing.
Here’s an idea of what that content looked like:
The full piece of content ended up at less than 400 words and got straight to the point.
Could we have written a longer, more comprehensive article? Yes, and our GPT articles usually are longer than this.
But we wanted to act fast and get this out immediately. And it turns out that this was all that was needed to re-train LLMs!
What happened: the GPT article started being cited within 24 hours
Here’s how I know this strategy worked.
Before we published anything, I tested the exact question we wanted to address across ChatGPT, Perplexity, Gemini and other LLMs.
“Does [our client] have the right licensing to operate in North America?”
The results were not great. Sometimes the competitor’s article was cited directly. Other times, the answer was vague and basically amounted to a “maybe”, so not exactly what you want a prospect to see when they are asking a pretty straightforward question about whether you can legally serve them.
After we published our GPT article, I ran the same tests again and started seeing our new page being cited in LLM responses, with the correct information now appearing in the answer.
We added the prompt into Peec.AI, the software we use to track the LLM queries we care about, and across the responses we monitored, that piece of content was now appearing 100% of the time for that prompt, and has been ever since.
All of this happened in less than 24 hours.
By publishing a clear, first-party source that directly answered a simple yes-or-no question, we gave LLMs a stronger source to work with almost immediately.
That won’t be the case for every GPT article. If you’re targeting a more subjective prompt like “What is the best X solution for Y?”, results will usually take longer. But for highly factual questions with a clear answer, we’ve found new information can surface incredibly quickly.
How to use GPT articles to answer the questions your sales team keeps hearing
Hopefully you don’t have competitors publishing false claims about your brand very often, but if it does happen (and we’ve had 2 other clients where something similar has happened) you now know what to do.
But what are some other ways you can apply this tactic?
The value here is that you can apply this tactic to any important question your prospects keep asking, especially factual ones you want to address quickly. The easiest way to find these questions is to talk to your sales and CS teams. Ask them what they’re constantly having to explain.
For example:
- How does your pricing work, and which features are included and which are add-ons?
- Which markets do you support? Do you offer all products in all markets? Are there limitations?
- How do you compare with a competitor?
- Do you integrate with a specific platform?
If prospects are asking your team these questions, they’re definitely asking LLMs too.
And if your website doesn’t answer them clearly, those LLMs have to work with whatever information they can find. That can easily be a competitor, comparison site or a random third party who aren’t motivated to paint you in the best light.
Use software like Peec.AI to track what’s being cited for the prompts, or just try running through LLMs yourself.
If you’re not happy with the answer LLMs are giving, create a GPT article that clearly sets out the accurate information you want prospects to find.
Using GPT articles to win in AI search
Correcting misinformation is only one use case for GPT articles.
The bigger opportunity is using them to grow visibility for the prompts your prospects are already searching, helping you turn LLMs into a true acquisition channel.
We’ve spent the last year working on improving AI visibility and leads with real fintech clients, and we’re sharing what we’ve learned in our upcoming report and webinar on How Fintechs Can Win in AI Search.
We’ll break down the exact 4 step AEO process we’ve used with fintech clients to grow LLM visibility and generate pipeline.
Sign up for the webinar and get the report straight to your inbox.
Mint Studios Recommended Reads
- Time to Citation in AI Search: How Long Until You Show Up?: An interesting blog on how fast most content appears on AI search.
- Case Study: How We Increased LLM Leads for Fiska by Over 360%: A recent case study on our client Fiska, where we were able to build a pipeline worth seven figures, in large part due to creating GPT articles.
- How ChatGPT, Perplexity, Gemini, and Claude Actually Decide What to Cite: A useful breakdown of what different LLMs tend to cite, based on analysis of millions of AI citations, and why having clear, trustworthy information about your brand is so important.
Thanks for reading,
Elliot & the Mint team 🎉
We help financial services and fintech companies turn content into measurable pipeline. Learn more about what we do.
