AI advertising · ChatGPT

Advertising where the question gets asked.

Your buyers now describe what they want to an assistant before they describe it to anyone else. We run the paid campaigns inside that conversation, and the earned work that gets you named in the answers. One team, one account owner, one readout.

Brief the team →
01What we run 02How it works 03The usual asks

What we run

What we actually run.

Two ways to show up in an AI answer, and the measurement that tells you which one paid for itself. Most brands want both. We run either on its own.

Paid · ChatGPT Ads

The card under the answer.

Campaign strategy, account build, creative and bidding on the OpenAI ads platform, run by people who do this every day.

  • Campaign and ad group architecture built around your goal
  • Creative written to the format, not resized from a display banner
  • Targeting written against real buying conversations
  • Budgets sized to produce a readable signal
Earned · AI visibility

Being named in it.

The work that decides whether an assistant puts you on the shortlist when nobody has paid for the answer.

  • Crawler access audited end to end, not just your robots rules
  • Pages restructured so an assistant can quote them cleanly
  • Entity and citation footprint across the sources models read
  • Product and pricing data published machine readable
Measurement

Proof it worked.

Wired before the first impression is bought, so the first report answers the question instead of raising it.

  • Pixel and server-side conversions installed and deduplicated
  • Conversion events agreed with you, not assumed
  • Prompt-level tracking of where you get named, and who instead
  • Reporting on the numbers your CRM already shows

Why it needs its own hands

It is not search. It is not social.

Two things break the habits a good paid team already has. They are the reason this does not run itself off the side of a search desk.

01 / Intent

Nobody types a keyword.

People explain the whole situation: the constraint, the budget, what they already tried. There is no keyword list to bid on and no demographic to stack. You describe the moments where you are genuinely useful, and the system matches the conversation. Written badly, it delivers nowhere. Written well, it reaches people mid-decision.

02 / The answer is not for sale

Paying does not buy the recommendation.

OpenAI's stated policy is that advertising does not influence the answers ChatGPT gives, and that ads are labelled and kept separate from them. So the shortlist inside the answer is earned, not bought. Anyone selling you a paid route into the recommendation itself is selling you something that does not exist. That is why both halves of this page exist.

How it works

How this works.

Four steps from brief to live, then a rhythm you can audit against. You never touch the platform unless you want to.

Step 1

Brief.

Send the goal, the budget and what you already have. We come back with a plan, the audience we think is worth reaching and a budget sized to produce a readable answer.

Step 2

Build.

Account, campaigns, creative and measurement built and wired. Your landing pages cleared for the crawlers that review them, which is where most campaigns quietly fail.

Step 3

Launch.

Spend goes live with a named owner. You see the same dashboard we do, from the first day it runs.

Step 4

Optimise.

Creative, targeting and budget worked against the conversion, and an honest verdict on whether the line earns its place next to the rest of your spend.

What you actually get

Five things you can hold us to.

Written down before the engagement starts rather than after, in the same terms as every other line we run for you.

  • →Creative and targeting written for this format, not ported across from a display campaign.
  • →Conversion measurement live before the first impression is bought.
  • →Landing pages cleared for the crawlers that review them, before anything is submitted.
  • →Reporting on the same numbers your CRM shows, not platform-only metrics.
  • →A named owner you can call, and a recommendation to stop if the data says stop.

Questions

The usual asks.

The things that come up in the first conversation. If yours is not here, put it in the brief below.

Can I pay to be one of ChatGPT's recommendations?
No, and be careful with anyone who tells you otherwise. OpenAI's policy is that advertising does not influence the answers ChatGPT gives, and that ads are labelled and kept separate from them. Getting named inside the answer is earned work. We run that too, and it is the half most people underestimate.
Do we have to choose between the paid and the earned side?
No, and most brands should not. The paid side buys reach now. The earned side compounds and decides whether you get named when nobody has paid for the answer. They use the same technical groundwork, so running them together costs less than running them apart. We are happy to start with either.
What does it cost?
Media is bought on impressions or on clicks, billed by the platform to your account at cost. Our fee is scoped against the work and quoted on your brief. We will also tell you the budget we think produces a readable answer, which is usually smaller than people expect for a first test.
How is this different from running Google or Meta?
There is no keyword list and no demographic or lookalike audience. You describe the situations where your product genuinely helps, and the system matches against what the person is actually working through. It rewards teams who can write about a customer problem plainly, and punishes anyone porting a display banner across.
Why would our ads get rejected or fail to serve?
Usually for a reason that has nothing to do with the creative. The landing page is crawled before an ad is approved, and firewalls, bot mitigation, browser challenges and rate limiting block that crawl far more often than people expect. We audit every layer before submission. It is also the same access that decides whether you can be cited organically, so it is worth fixing either way.
Does this work in our market?
The paid platform has opened market by market rather than everywhere at once, so that is one of the first things we confirm for you. Where it is open we run it. Where it is not open yet we run the earned side, which works anywhere, and pick the paid side up the moment it arrives. You are not left waiting either way.
Can you run this alongside the rest of our media?
Yes, and that is usually the point. We plan and buy across audio, video, press, out of home and digital, so this sits in the same plan and the same reporting rather than in a separate tool nobody opens.

Brief the team

Send us the brief.

Who you are trying to reach, what you want to happen and what you have already got. We will come back with a plan and a name on it, or with the reason not to bother.

What's this about? Pick one to sharpen the routing
Markets Pick all that apply
Budget
Timing
We'll route this to
Pick a medium above and we'll route this to the right desk…
→
We never share your details, and you can ask us to delete the brief at any time.

Adjacent practices

Also worth a look.

Three practices run by the same team, with common framing and common reporting.

Practice · Organic

Authority OS.

Technical SEO, content strategy and editorial production aimed at compounding organic traffic. The foundation the earned route builds on.

View practice → Practice · Paid search

Paid search.

Search, shopping and display managed against revenue events. Where the demand you build in an AI answer usually converts.

View practice → Practice · Digital

Digital hub.

Every performance practice we run across search and social, under one account team and one readout.

View practice →