Like Your Brand in AI, This Post Comes In Pieces

Imagine clicking through your morning emails. You’ve recently made some budget decisions for AI, based on what staff and vendors have been telling you. And it all strikes you as the AI version of SEO: dashboards with counts of your brand in AI responses and sources AI cited. Visibility box ticked.

Then, on your morning scroll, you see a subject line: What version of your brand will AI show today?

That’s an uncomfortable question, right up there with: Why would a machine recommend your brand? Many brands have a good answer, and yet AI doesn’t name them consistently or correctly.

Your brand is very likely scattered across many of the sources AI draws on to answer a customer’s question about your category. Your website is an important one. So is a government registry, maybe an industry directory. How about that “best of” list from 3 years ago, a COVID-era blog post, the one with a photo of everyone in face masks, or the Wikipedia page nobody has looked at since the last acquisition? AI may be pulling on those old sources too, to decide if you belong in a response about your category today.

Your brand has changed over the years, but AI may not know it. Or if you are new, AI may not know yet why you matter. AI encounters information of widely varying quality, written at different times, for different purposes. What it draws from that information can determine the version of your brand a customer sees.

One source may have your current name, another a press page about a product announcement or category you left years ago. Still another about a new partnership that foretells an exciting future (and product launch). Each is (or was) correct at the time, but together, can lead AI to describe a company in a way that conflicts with its current positioning. Finding those pieces, then understanding where they agree and where they conflict, is the beginning of putting the brand back together. While you can’t control AI, you can see the sources AI uses to define your category and your organization. Once you see them, you can begin to manage them – and measurement is the key.

For the past year I’ve been building tools that let me see brands in AI systems. For the last four months I’ve been using those tools on real brands. Let me tell you what I see. It is a detective story. For example, I’ve seen a large, well-known, and trusted brand rarely appear in AI responses to questions about services it has provided for more than 100 years, and discovered the likely reasons, along with where to start making changes.

The goal is a picture of how AI represents your category, your competitors, and your brand, along with the sources it cites. We can’t know everything an AI system relies on to produce an answer, but by repeatedly measuring responses and their citations, we can identify the sources worth investigating.

This is a short series of articles about how that works.

The customer’s question

I am guided by what a curious customer might ask about a category – questions that don’t name any one brand. The client in our detective story has their own questions: Are we in that answer? How are we described when we are? Which sources did AI cite, and what do they say about us?

To answer the client’s questions, I ask AI a fixed set of customer questions, many times, on 4 AI systems (so far), and read every answer. Was the client named, where on the list, how is it described, and are the facts right? Wrong facts are just as important. AI will confidently recommend an agency that closed years ago or list an organization by a name long since retired.

The third client question is about sources. I log every source of information AI cites in its responses – about 1,900 in my last audit – and open each one. There is a lot of noise in those sources. Most are a news article, a competitor’s own website, or (often) a page about something irrelevant. Opening and checking them is the only way to go from noise to a working list of what matters.

What you get

A detailed picture of how AI represents your brand in the category, followed by a prioritized list of sources that help explain it. The industry registers, directories, and top publications that carry (or wrongly omit) the client are the list to be pitched for corrections or inclusion. Some changes can only be earned or made indirectly. Others involve slow and uncertain correction processes (I’m looking at you, Wikipedia). Some can’t be influenced at all, of course, but I’m an optimist, and that list of sources and recommended actions is where measurement meets your marketing team (or agencies).

Then next quarter I do it again, the same way. Back to the action list, item by item: what was done, and did the AI notice?

Isn’t this just SEO for AI?

Or GEO or PR? Those are important capabilities, but they come later in the process and are hopefully well covered by your existing team and vendors. My role is measurement, and it comes first, followed by source investigation and recommendations for changes that matter most to the business and its brand. It’s how a decision-maker can see their brand, competition, and category as AI sees them, and establish a process for accountability.

Yes, the uncertainty can be frustrating, which is why frameworks like the IAB’s, “Measuring Visibility in the AI Era” are so useful: common standards for measuring whether a brand appears in AI responses, how prominently, and how it is portrayed or recommended. The real work is investigating the sources cited in those responses, finding what’s missing or wrong, and which of it matters, to give the brand’s marketing team the evidence it needs to do something about it.

In the next post, I’ll explain how I find the pieces.