Last week, the chatbot conversation picked up two business models at once.

On Monday, Digiday reported that ChatGPT is about to show brands their mention count. A day later, it reported that OpenAI’s new business agents give brands a free way into the conversation it’s also selling as an ad. The same day, AdExchanger’s headline put the pecking order bluntly: in OpenAI’s ad business, advertisers come third.

Read together, those three stories describe a channel that is a media property and a storefront in the same window. A brand can pay to appear in the conversation, show up for free through its own agent, and be measured on how often the assistant brings it up unprompted. That’s new, and it changes what a marketing team has to be good at.

Everyone is building the counter

OpenAI wasn’t alone. AdExchanger’s Wednesday roundup was headlined Google’s AI ad losses are Meta’s gain; chatbot ad formats are starting to multiply. MediaPost reported that Criteo’s chief predicts AI platforms will adopt advertising, which is a safe prediction from a company whose business depends on commerce ads having somewhere to run.

On the storefront side, BMW launched a ChatGPT shopping plug-in, and Pinterest’s new AI assistant offers product conversations and comparisons. Google, meanwhile, is working out how to monetize visual to voice search without media buys, a headline that should make every search marketer sit up. If the platform can earn money from the answer without a media buy, the advertiser’s budget is no longer the only way in.

That’s the core tension in AdExchanger’s “advertisers come third” framing. The assistant’s first job is to be useful to the person asking. Its second is to keep the platform growing. Ads fit in after that. Brands that treat chatbot ads like a new row on the media plan will be buying into a system that is designed, openly, not to put them first.

We’ve seen the assistant pitch before

The archive has two earlier rounds of this idea, and both are useful.

In 2016 and 2017, brands built chatbots of their own. Digiday covered Burberry’s “see-now-buy-now” chatbot, and fashion and retail brands lined up behind it on Facebook Messenger. Most of those bots were brand-owned novelty. They lived inside a messaging app, answered a narrow set of questions and quietly disappeared.

Then came voice. In 2018, voice search and voice assistants came up in roughly one of every 100 stories we collected. Voice shopping forecasts were enormous, even though Digiday had warned brands a year earlier to beware of the hype that surrounds voice search advertising. By 2021 the topic had fallen to about one story in 700. We’ve written about that curve before.

This round is different in one important way. The 2016 bots belonged to brands, and the voice assistants never built a real ad business. This time the assistant belongs to the platform, the platform is selling ads, and consumers are actually using it to shop. Chatbots and ChatGPT went from one in every 450 stories in 2017 to about one in every 65 this year. The closer comparison isn’t the chatbot fad. It’s the early days of search, when brands had to learn to buy keywords and earn rankings at the same time.

Two jobs, two audiences

That search comparison is the right frame, with a twist. In search, paid and organic sat side by side on a results page, and the brand could see both. In a conversation, the assistant decides what to say, and often says only one thing.

MediaPost put it plainly for business marketers: B2B marketers must now win two audiences: human, machine. A MarTech contributor from Optimove wrote about connecting with shoppers when the buyer is not human. The measurement tools are arriving to match: ExchangeWire reported that MarkApp launched Shelf, AI visibility analytics that measures a brand’s share of the AI answer, and ChatGPT’s own mention count is on the way.

“Share of the AI answer” is a useful phrase, because it names the second job. The first job is buying the placement. The second is being the answer the assistant gives when nobody has paid for anything. Those are different skills, owned by different people in most organizations, and the second one depends on things media teams don’t control: product data, reviews, pricing, availability and how clearly a brand explains what it sells.

The people at the top haven’t caught up. MediaPost’s headline last week was that CMOs are not keeping pace with the rise of AI assistants. We’d guess the problem is less awareness than organization. Nobody owns “the answer.”

We covered the buying and selling side of this in our piece on agentic advertising. This is the other half: what a brand has to look like when the thing reading it is software.

What we’d do now

Give the answer an owner. Someone needs to be accountable for how assistants describe your brand, products and prices. It sits somewhere between SEO, product marketing and e-commerce. Pick one person.

Measure before you buy. Mention counts and share-of-answer tools are early and imperfect, but they’re cheaper than a test budget. Find out what assistants already say about you, and where they’re wrong.

Fix the product data. An assistant comparing options can only use what it can read. Current, structured product information, kept in a CDP or cloud data platform and published consistently, is the closest thing to SEO hygiene in this channel.

Treat the free door as seriously as the paid one. OpenAI is offering business agents at no cost while it sells ads in the same conversation. Use both, and keep the paid side honest by tracking what you’d get without it.

Buy chatbot ads as tests, not as a channel. The formats are multiplying and the rules are being written by platforms that say openly that advertisers aren’t first. Small, measured tests will teach you more than a big commitment.


Every headline in this piece links to the original story. Browse the archive’s coverage of ChatGPT ads, chatbots and voice search.

Kevin Reynolds is the founder of Advertising Perspectives, MadTech Consulting and data4orce.ai. He is a marketing technology architect and C-level advisor, with a focus on Salesforce and data-driven advertising.