Reputation in AI-first discovery: how does your brand show up?

More than half of B2B buyers now start their research journey with an AI chatbot rather than a search bar. G2's Answer Economy Report put it at 51% in 2026, up from 29% a year earlier, and a markedly high 69% said the answers they got changed the vendor they had intended to choose. The machine did not just inform the shortlist. It rewrote it.

So what does this mean in practice? Well, when someone asks ChatGPT, Gemini, or Perplexity who the leading players in your category are, the model does not hand over a page of links and call it quits. It answers. It reads everything ever written about you and your competitors and hands your buyer a verdict they rarely question. Your first impression is now an answer to a prompt you never get to see.

The instinct, when inputs multiply like this, is to grab for more control, but in this instance that is fundamentally wrong. You were never fully in control to begin with, you had fewer sources to watch. So the real question is not whether you rank. It is who owns how your brand shows up in those answers. There are two halves to this: are you in the answer at all, and is what it says about you true. Miss either and the buyer is gone before you knew they existed.

Maybe I would say this as a communications practitioner, but the honest problem is that everyone has a stake and no one owns it. Paid, SEO, content, and comms all shape the answer, yet each sits in a different budget line, so it falls through the gap. Someone has to own the story, and communications is the natural home for it: pulling the others in, not pushing them out. Showing up well in AI search is not a technical problem, it is a reputation problem, and reputation is what comms teams have always managed.

What is Generative Engine Optimisation?

The discipline forming around this has a name: generative engine optimisation, or GEO. If SEO was about earning a place on a page of links, GEO is about earning a place in the answer itself. Call it PR generative engine optimisation, because that is what it is: reputation management for an audience of machines. And although you may only just be starting to get your head around the concept, it is already dramatically changing your relationship with your key audiences. Forrester's 2026 Buyers' Journey Survey of 18,000 buyers found 94% used AI in their most recent purchase, and answer engines now outrank vendor websites and sales reps as the top research source. Comms AI search is the shiny new front door, and most brands are stuck re-painting the old one.

To see why this sits with comms, look at how a model forms its views. They read across every department at once: marketing's landing pages, PR's coverage, HR's Glassdoor page, product's pricing page. Then it hands the buyer a single story. You cannot tell that story well with four owners in four divisions, and the buyer never sees your org chart anyway. The fix is not picking a winner among them, it is giving one team ownership of the narrative and having the rest feed it. Smaller, more agile businesses have an advantage: one person already owns the story.

This is where the difference between manufactured and earned reputation stops being theoretical and starts costing you deals. Publish a hundred keyword-stuffed pages claiming you are the category leader, and the model weighs them against what credible, independent sources actually say and sides with the evidence. The only reputation that survives being cross-referenced against everything ever written about you is an earned one.

Measuring reputation in AI-first discovery

The work is less exotic than the acronyms suggest, and starts with measuring honestly. Ask the AI tools the questions your buyers actually ask, the flattering and the brutal. What comes back tends to fall into four jobs.

1. First, the blunt one: does it mention you at all? Absence is an owned and earned content gap, not a correction. You need enough credible content in the sources these models trust to show up in the first place.

2. Answers that are true but uncomfortable need a positioning fix: own the framing before the model does.

3. Outdated answers need earned media: you do not correct an old narrative, you outweigh it with stronger coverage.

4. Wrong answers need correcting at source, in the places the models trust.

It is why we built Surfacd: to see how AI describes a brand, which sources it draws on, and where the answers drift from the truth. Notice none of that is more press releases. The work has moved up a level, to framing, evidence, and sources.

None of this is a reason to panic. AI-first discovery rewards the brands that have been doing reputation properly all along: substance over noise, consistency over campaigns, and evidence and proof points over fluffy adjectives. It removes the hiding places. It is a compounding asset, not a sprint: you cannot switch it on the week before a raise, a deal, or a crisis. Instead, you have to be the tortoise: slow, steady, and always there, with consistency of narrative beating speed every time.

The old question was whether buyers could find you. The question now is this: when the machine speaks for your brand, does it mention you at all, does it tell the truth about you, and is that truth good enough to win? That is the question worth building your reputation around.