ChatGPT Ads are sponsored placements shown inside an AI assistant conversation when a person asks something commercial. They sit alongside the assistant's own answer, so the ad is competing with a direct, well-written response for attention. That one fact changes how targeting, creative and landing pages have to work. This post walks through each. The service page is ChatGPT Ads by MediaBloom.
Targeting: intent, not keywords
Search ads match the words a person typed. Assistant placements match the meaning of the prompt and the context of the conversation around it. "We miss a lot of calls after 5pm and lose jobs" contains no keyword, but it is a buyer describing the problem a 24/7 AI receptionist solves. Campaigns are therefore built around prompt clusters: the questions and problem statements buyers actually type, grouped by the offer that answers them. The best source for those is already in your Search Console: the long, prompt-shaped queries that AI systems send when they retrieve your pages.
Creative: write an answer
An ad that reads like a billboard loses to the assistant's answer every time. The ad that wins reads like the most useful next sentence in the conversation: specific to the prompt, with one proof point and one clear reason to click. "24/7 AI receptionist for home services. Answers every call, books the job, texts the customer. See a recorded call." is closer to it than any headline about being the leading provider of anything.
Landing pages: finish the conversation
The visitor arrives with a question half answered and a decision half made. The page has to answer the rest in its first screen: what it is, who it is for, what it costs or how pricing works, and proof. Then the next step, booking a call or getting a quote, in one click. Pages written for search traffic, with a hero slogan and a scroll of features, waste the intent the click arrived with. We build these pages the same way we build answer-first pages for AEO.
Tracking: prompt to appointment
Click tracking is table stakes. The number you need is cost per booked appointment by prompt cluster, which means the lead has to carry its source into the CRM and the appointment has to be attributed back. We set this up in GoHighLevel, HubSpot or Salesforce before launch, as part of our CRM automation work, so the first weekly report already shows which clusters book.
Follow-up: the part that decides the cost
An ad click is a lead for about five minutes. If the business calls back in that window the lead converts at a multiple of the rate it would an hour later. That is why every lead from these campaigns is handed to the speed-to-lead AI setter the second it arrives, by text and voice, with the appointment booked onto a rep's calendar. The same campaign with slow follow-up can cost two or three times as much per appointment.
Setup order
Scoping and prompt research first. Then tracking and landing pages, because launching without them produces spend you cannot read. Then creative per cluster. Then a test budget for two to four weeks, judged on cost per appointment. Then scale the clusters that book and cut the ones that do not. For what to budget, read how much ChatGPT Ads cost.
TODO(John): Add a screenshot or anonymised example of a live ChatGPT Ads placement and its landing page from a MediaBloom campaign.
