Case study · ChatGPT Ads

ChatGPT Ads for a B2B offer: from eligibility check to lead in HubSpot

Since August 2026, businesses in Germany can run ads in ChatGPT. For a B2B service provider we built the channel from the ground up – from the eligibility check through a dedicated landing page and measurement after consent to the lead handover to HubSpot. It is built as a system: every further campaign, landing page variant, ad or report builds on the same tested components instead of starting from scratch. Nothing is activated or changed without a human’s approval.

Armin StadlerAs of: Deutsche Version

Starting point

The client sells a service to businesses that needs some explaining. What they wanted was an additional channel for qualified enquiries: not reach and not clicks for their own sake, but completed forms from companies that fit the offer.

ChatGPT Ads works differently from search advertising. There are no search terms to bid on, no personalised ads in the European Economic Area and no benchmarks to orient by. And one question was open before launch: in the initial phase, OpenAI’s advertising policies focus on consumer categories – household and lifestyle, local services, travel, digital products and education. A B2B service fits none of them clearly.

What we built

Building blocks that together form a complete lead channel – from the first check to the record in the CRM. And a system every further campaign builds on.

An eligibility check before the first euro

We checked the offer against the current policies and named the approval risk openly. The decision was deliberate: one truthful submission, and no workaround variants if it were rejected.

A campaign built on situations, not search terms

In ChatGPT Ads, so-called context hints describe the conversational situations in which an offer is useful. We derived them from the typical situations of the target group and kept the campaign deliberately lean, so a test budget is not spread across too many ad groups.

A landing page with a qualifying form

A dedicated, fast-loading landing page with a second variant for testing, open to OpenAI’s review crawler. The form asks for role, company size and area of interest – so sales knows who they are talking to before the first call.

Measurement after consent, leads straight into HubSpot

OpenAI’s measurement script only loads once visitors agree. Every enquiry is validated on the server and handed straight to HubSpot. Only once HubSpot has accepted the lead – and only with consent – does the page report the conversion to OpenAI, via pixel and server with a shared event ID so nothing is counted twice.

A system, not a one-off campaign

Everything is set up to be repeated and extended. New landing page variants are built on the same tested layout, form and handover path. New campaigns and ads are prepared as a plan file, automatically checked against the character limits and approved via a checksum. Platform knowledge is kept with source and check date, and automated tests cover every component. That turns every further campaign into a routine step instead of a rebuild.

How the project ran

  1. 01

    Product brief and sources

    Offer, audience, language rules and statements that are off limits, written down. Every statement about the platform backed by an official OpenAI source and a check date.

  2. 02

    Eligibility and risk

    Eligibility checked, the risk documented and the submission strategy approved by the client.

  3. 03

    Landing page and lead path

    Page, form, consent and HubSpot handover built, and tested end to end with a controlled test lead before launch.

  4. 04

    Create the campaign – paused

    The campaign plan exists as a file and is approved via a checksum. Everything is created in paused status.

  5. 05

    Joint review and launch

    Ads, budget and run time reviewed together with the client in Ads Manager. A human activated the campaign – not the system.

  6. 06

    Operations and evaluation

    Every later change needs its own approval and a check against the live state afterwards. Success is judged by the quality of enquiries in the CRM, not by early click-through rates.

Where AI helped – and where it did not

We delivered the project with an AI agent for development and research. The line is drawn clearly:

What the AI took on

  • Researching platform rules – with source and check date
  • Drafting ads and context hints, automatically checked against the character limits
  • Code for the landing page, form validation and HubSpot handover, covered by automated tests
  • Read-only analysis of the ad account

What a human decides

  • Whether to take the submission risk
  • Budget and run time
  • Activating the campaign
  • Every change to live ads

Result

All submitted ads were approved by OpenAI, and the campaign launched in September 2026. That is a single review outcome, not a promise for future campaigns: OpenAI can change its policies and decisions at any time.

The lead path is in place: every enquiry arrives in HubSpot validated and tagged with its landing page variant. The bigger gain, though, is in everything that comes after. Further campaigns, ads, landing page variants and reports build on tested components instead of being created from scratch each time – which saves set-up work, keeps the risk of errors low and makes the channel extensible at any time.

We deliberately do not publish campaign figures. The platform is new, OpenAI itself publishes no benchmarks, and the data belongs to the client.

Prepared for the next step

Prepared

The next stages are already in the code and are switched on as soon as they are needed. Even then, a human approves whatever goes live.

  • Automatic weekly reports via read-only access to the ads interface
  • New campaigns, ad groups and ads straight from approved plan files – always created paused
  • Further landing page variants on the same tested form and handover path

What carries over to your business

  • A setup that needs less building with every further campaign, because every component is reusable.
  • The up-front eligibility check – before time goes into a setup that will never be approved.
  • A lead path that measures nothing without consent and still gets every enquiry cleanly into the CRM.
  • Approvals at fixed points: submission, budget, activation and every change.
ChatGPT AdsLanding pages

Frequently asked questions

Can a B2B offer advertise in ChatGPT?

In the initial phase, OpenAI’s advertising policies focus on consumer categories; other categories are not generally foreseen at launch. In this project the ads were approved after a truthful submission. That is a single outcome, not a rule – which is why we check every offer against the current policies first.

Why don’t you share click or lead numbers?

Because figures from a single account on a new platform are easily mistaken for benchmarks – and because they belong to the client. OpenAI itself does not yet publish benchmarks for ChatGPT Ads across advertisers or industries.

Is anything measured without consent?

No. The measurement script only loads after consent, and only then is a conversion reported to OpenAI. The enquiry itself still reaches HubSpot: measurement becomes less complete, but sales does not lose a lead.

Why a dedicated landing page instead of the existing website?

OpenAI reviews every landing page before approval and uses its content to assess relevance. A dedicated page can be tailored exactly to the ad, opened to the review crawler and changed quickly – without touching the existing website.

What does the setup bring for further campaigns?

Most of the work goes in once: eligibility check, landing page path, measurement and CRM handover. Further campaigns, ads and variants build on that – prepared as a plan file, automatically checked against the character limits and approved via a checksum. New landing page variants reuse the same tested path.

Next step

Want to build something similar?

In a first call we work out whether the approach carries over to your business – and what it would take.

Book a first callChatGPT Ads

More case studies

All case studies

More on the blog