ChatGPT Ads for B2B Software: 490 Million Weekly Messages and What a Client Account Shows

OpenAI's September 2026 stat pack counts over 490 million ChatGPT messages a week categorised as B2B or B2B-adjacent, averaged across 6 to 26 July 2026, plus a further 90 million on core small-business operations that OpenAI itself labels "not a B2B buyer-intent measure". Two of its six conversion case studies are software and technology advertisers. And we have a client account in the category with a few weeks of data. This is what the three sources add up to.

The demand number, read carefully

490 million is the weekly average of global messages that OpenAI's classifier assigned to a B2B topic. It is the smallest of the big categories in the pack, a fifth of professional services and a third of retail, and it is still the largest pool of business-buying conversations that any single platform has published. The business-operations figure sits next to it as a warning label: OpenAI went out of its way to say that "how do I register for VAT" is not a buying signal, and the same discipline applies to reading the B2B number.

What makes B2B different from retail on this channel is the length of the journey. A business buyer asks ChatGPT to explain a category, list options, compare two vendors and draft the internal case, often across several sessions. The pilot deck's own split put problem framing, discovery and comparison at around 58% of shopping conversations, and in B2B that share is larger. The ad sits under the answer at each of those steps, in front of the person who will write the shortlist.

The two published software cases

Both software and technology advertisers in OpenAI's pack moved most of their spend to Conversion Optimization and compared the 14 days after with the 14 days before. The first shifted 98% of spend and saw click-through rate rise 27% and cost per order fall 50%. The second shifted 73% and saw click-through rate rise 22% and cost per order fall 87%. Both measured through the Conversions API, both are labelled observational, and both are consistent with the median 44% CPA reduction OpenAI reports across 27 advertisers. The eligible events OpenAI lists include leads, registrations, trials and subscriptions, which is the whole B2B funnel. The named advertisers in the same pack point the same way: HubSpot reports competitive CPAs across a number of campaigns and downstream revenue signals, and Figma says ChatGPT Ads has become a top performer for its Figma Make campaigns, with higher click-through rates than traditional search.

What a client account shows

We run ChatGPT Ads for a B2B software client in the United Kingdom, whose name stays confidential. Over the first weeks on click bidding, cost per click fell about 29% as the platform learned which conversations fit the ads, with click-through rate holding in a range of 0.8 to 0.9%. Clicks have cleared between roughly £1.50 and £2.30, below OpenAI's $3 to $5 guidance. The account is on click bidding because its primary event, a booked demo, has not yet reached the five click-attributed conversions that OpenAI's own benchmark used as its floor; the step before it, a tap on the booking link, is the event accumulating history. That sequence, pixel, Conversions API, five conversions, then the switch, is the one every published case followed, and it is the one we are following.

How a B2B campaign is built

Context hints describe the buyer's situation. For a workflow-automation product, as an example: an operations lead evaluating tools to replace spreadsheet handoffs; a founder asking what an AI agent can do for customer support; a team comparing two named platforms; someone drafting a business case for automation budget. Each becomes an ad group with its own creative and landing page, and within a month the account reports which situations the market actually asks about, which is a demand map as well as a media plan. The Conversion Optimization guide covers the switch, and the ad generator writes the first hundred ads against the real character limits.