Inhuman

The biggest thing happening in advertising, isn't what you think.

For the first time ever, your brand has a non-human audience.

This is the potentially biggest strategic shift in the history of marketing, and 99.9% of conversation about how AI is going to change advertising seems to be about whether budgets are going to decrease.

This is the biggest moment since the MBAs convinced us to separate media and creative, and we’re about to make a similarly dumb mistake. We’re solely looking at AI in terms of efficiency and staffing and consumer search behaviour, and forgetting that it’s a brand new target that is going to be ingesting what you put into the world, and adding it to the understanding that more and more humans are starting to rely on every day. Potentially the biggest decision-making influence for the next generation, and we’re worried it’s gonna take our jobs, instead of being worried it’ll just make them irrelevant.

A slide from a Field Trip deck presented at FUSE Create on June 4th.

We’ve got to at least consider marketing to the machines. AI is an audience, now. Here’s some very theoretical ideas about what that could mean.

Reinventing Messaging & Strategy

Everything you’ve ever done to develop a messaging hierarchy or a communications strategy is probably obsolete. I know, that’s a lot for a first subhead. Sorry. But we’ve built all of these things on a semi-linear understanding of storytelling, an exclusively human lens on retention and motivations, and the idea that it’s all inherently somewhat temporary. But when there’s a handful of all-knowing all-remembering models that are interacting with a growing segment of the world, and mediating their connection to established and emerging information, the stakes are little bit higher, and the assumption of a context window based on anyone you know, is inherently a little risky. I will be transparent that I’m just starting to wrestle with how this is going to change, but I am nearly certain it will have to - we’re training non-human systems on information built for humans, and it feels a lot like expecting someone who has just learned English to be fluent in the specific idiom of your small section of your home country.

Share of Training Data

Share of Voice is obvious and valuable. If the correct amount of messaging in a category comes from your brand, it maps to the volume you’ll sell, how top of mind you are, and general perception of your brand. But when your audience is being consistently trained on more and more up to date sets of data, as well as scraping new information on the internet, I’m wondering if Share of Training Data isn’t going to become equally important. Because there’s no public listing, you have to assume available volume of valuable content is going to have an influence on how often your brand shows up in training data, and how ‘top of model’ a specific brand is. This doesn’t mean having the world’s largest website, but it DOES mean thought leadership, white papers, and valuable branded research now have a secondary purpose. It probably also means rethinking the signup sheet that you’re putting in front of content in hopes of list building, if you’re a B2B or B2C brand. What ways can you create valuable training data that is both neutral/trustworthy, and brand supportive? You probably wouldn’t do this entirely separately from your existing marketing activity, but you might consider it a valuable discussion to attach to what you’re already doing.

Rethinking Long Form

The attention span problem isn’t a problem anymore… or at least it won’t be for long. Many people I know are already turning to AI habitually to get the summary of reports or essays they don’t have the time or inclination to actually read (this is a way of avoiding drudgery, all of them still read things they’re interested in, you don't need to comment outrage, all good) and a few edge cases are using AI to distill and transcribe the key points of video content. Suddenly, long form is both a way of expressing nuance and depth, but it’s also a way of shaping the machine. People can engage with long form as they see fit; a summary, skipping to specific chapters, and hopefully as a buffet of images or ideas they can peruse at their leisure if interested. What’s interesting is that the long form, with no cuts based on assumptions and with maximum detail for nuance, is going to be way better for an AI audience, and valuable and appreciated for a hyper-engaged audience. The argument against making longer form content (when it makes sense) is probably getting weaker.

Rethinking Social and UGC

We all like Social & UGC. Positive impressions and feedback and reach and community are all key, but this there’s a decent chance UGC is also going to end up shaping what a major AI platform thinks about your brand or product. Suddenly incentivizing UGC isn’t just valuable for the short term life of the campaign or promotion, but incrementally as an increase in the aforementioned SOTD metric. But think about what this means in terms of PR, of sentiment, of social engagement. These interactions are often treated as spray & pray - low cost, a place where you can take a lot of shots, a place where high risk high reward ideas can sometimes deliver the case study of a lifetime. What does it mean when there’s a decent chance all of it is being recorded somewhere and will influence how an influential system thinks and talks about your brand? If you run a contest and it leads to online backlash, is that going to meaningfully shift perception when someone asks Claude for a POV? I have no idea, but I have theories.

Explicit Salience

I’ve been obsessing lately over the idea of a “contextual brand”, one developed with a distinct understanding that it needs to make sense in a broad range of channels, situations and formats, and that making sense can be paired with some level of tailoring - LEGO shows up well in animated comedy, in fake desk plants for nerds, in children’s toys, and in pop culture. It all feels different, but all authentically LEGO. What hit me, is this won’t translate to a non-human audience. We’d need explicit salience in each of these contexts, baked into the way we speak about the brand, on record. We have to be explicit because we cannot assume any cultural context, nuance, or allusion is going to come through. Brand working without implicit associations is not brand as we have conceived of it. And so we need to vastly re-imagine how we build and communicate about brands in a reality where a human cultural understand is not a given.

These are shots in the dark, to be clear. But I think they’re relevant ones.

After writing all of this, I asked Perplexity. This is the response I got:

Key Principles for Marketing to LLMs

• LLMs prioritize resolution, not attention: Unlike humans, LLMs are not swayed by emotional appeals or flashy branding. They are designed to resolve user queries with accurate, relevant, and trustworthy information.

• Visibility is measured by Share of Model (SOM): This new metric assesses how often your brand is mentioned, the gap between human and AI awareness, and the sentiment associated with your brand in LLM outputs.

• Content must be structured and context-rich: LLMs draw from a wide range of sources—text, images, videos, reviews, and structured data. They favor content that is well-organized, factual, and directly addresses user needs or queries.

We’re going to need to figure out how to market to the LLMs, as much as we’re figuring out how to market with them. Nearly everyone I know in this industry is obsessing over the latter, and not thinking about the former at all. I’m starting to think that’ll be a determining factor in how the next decade goes.