What Is AI Saying About Your Brand?

A year ago, I asked ChatGPT what it thought about HubSpot. Not because I was trying to be clever. I wrote about it in this newsletter at the time.

I'd discovered an agent on agent.ai (built by Paul Friberg) that compared brands through ChatGPT's eyes, so I plugged in HubSpot vs. Salesforce and read what came back.

The results were more interesting than I expected.

ChatGPT had a well-nuanced read on our strengths. But it also had specific blind spots: it thought our customization was limited, and it underestimated how many integrations we actually ship.

That was a year ago.

Best practices around Answer Engine Optimization (AEO) have since evolved -- and yet many otherwise advanced AI practitioners are still missing what I believe is the most important step.

Wait, let’s take half a step-back. You may not yet be familiar with AEO. Here it is in a nutshell. Way back in pre-historic times (like about 25 years ago), we had these things called search engines, the most popular being Google. If you had a question or wanted to find something, you typed it into the Google search box and it gave you some blue links to web pages that were likely to answer your question. The practice of optimizing a web page so that it showed up for specific searches in Google’s results was known as Search Engine Optimization (because Google is a search engine).

Fast forward 25 years and now increasingly, when we have a question, we ask AI tools like ChatGPT, Claude and Perplexity. Instead of search engines, we can think of them as answer engines. Instead of getting 10 blue links, these tools often just give you the answer (often with some citations to how it got that answer). And, just like we had SEO to optimize for search engines, we now have a practice known as AEO to optimize for Answer Engines (powered by AI). Now, instead of wanting your web page to be listed in the “10 blue links” of Google results, you want your company/product/offering to be what the answer engines tell users.

Example, if I ask ChatGPT or Perplexity “What are the top 3 CRMs for SMBs?” the answer will usually have HubSpot on the list (I just tried it, and it was #1 in both cases).

So today, I want to break down:

• How to create content for LLMs

• The biggest AEO blind spots

• How to monitor your AEO performance

—@dharmesh

How to Think About LLM-Friendly Content

The atomic unit has shrunk.

For the entire SEO era, the unit of the internet was the web page: Google indexed pages, the link graph ranked pages, and people clicked pages. When you optimized, you optimized at the page level, and the whole page either ranked or didn't.

In the AEO era, this is no longer the case.

The LLM pulls a chunk of your content, synthesizes whatever it found, and hands the human an answer. The unit that gets cited is smaller than the page -- it might be a paragraph, or the specific line that most clearly answers the user's question.

This suggests a new way to think about your content. You can keep writing long-form articles (and you probably should) but it helps to also think in chunks:

• What question is each section answering?

• Could you answer that question more precisely?

• Is that answer capable of standing on its own?

A high-quality sentence, to LLMs, is one that's easy to quote: snackable, pithy, precise enough to fit cleanly inside the answer the model is about to hand the human.

A 400-word paragraph that sort of gets at the question is much less likely to be cited than a tight sentence that delivers exactly the answer.

Think of it this way:

If the answer is buried inside a meandering narrative, the model has to work to find it.

If it's clearly marked and cleanly phrased, the model can hand it to the human with almost no editorial intervention.

And oddly enough, AI apps are a bit like humans. Not that they’re lazy per se, but they try to be efficient.

So, think in chunks, make it easy for the agent to do its job, and you may be rewarded for it.

The AEO Blind Spots Being Ignored

Tracking AEO visibility is an easy win.

That said, a year into the AEO conversation, the number of companies that are actively monitoring how AI talks about them is still very small.

Most have read about it, talked about it, maybe assigned someone to think about it.

Very few have actually run the comparisons: their brand against their three closest competitors, across the answer engines that matter, for the prompts that have the highest potential to lead to conversions. And by “prompts”, I mean what are humans likely to type into the AI app when looking for what you have to offer? (It’s the modern equivalent of the search term/keyword you’d type into Google).

Without tracking, how can you learn which prompts actually matter to your business?

Writing content that addresses the questions your audience is asking is good. Restructuring your existing top pages to be more LLM-friendly is great.

But how can you even validate the effectiveness of these tactics if you’re not monitoring your AEO rankings to begin with?

What gets measured gets managed. There are 3 blind spots I believe to be rather important when it comes to AEO:

1. Visibility. Am I showing up at all in the impactful answers?

2. Competitor share-of-voice. If I’m not showing up, who is?

3. Citation quantity and quality. What’s feeding these answers?

You can't optimize what you can't see -- step one of AEO is to address the visibility problem.

How to Gain Visibility Into What’s Working

Last year, "asking the AI" mostly meant asking ChatGPT. Brand visibility on AI was effectively a single-channel challenge.

That's no longer the case. ChatGPT is still the volume leader, with around 900 million weekly users, but Gemini and Perplexity are legitimate answer surfaces in their own right.

Tracking AEO performance has become a multi-channel effort.

If you want to see how AI is actually describing your brand today -- across ChatGPT, Gemini, and Perplexity -- that's what we built HubSpot AEO to do. It tracks your brand visibility across ChatGPT, Gemini, and Perplexity, compares you to competitors, and gives you a clear action plan.

I might be a bit biased since it’s built by HubSpot, but I think it’s a great way to actually start using AEO. If you’re a Pro/Enterprise user of HubSpot, it’s already included. If not, it’s only $50/month. One of my favorite parts of my job as CTO of HubSpot is being able to be involved in building the tools I wish I had for myself.

AEO is still very early (and so is HubSpot’s product) -- but not so early that it can’t bring you value. It’s very, very real. For many companies operating in mainstream industries, even a modest investment in some basic AEO usually yields measurable results (and I do mean measurable -- you can actually see the traffic from ChatGPT/Claude/etc. go up).

The companies that move on this in the next twelve months will look, in retrospect, the way the early SEO winners did.

—Dharmesh (@dharmesh)