ChatGPT Ads campaign management: writing context hints and fixing delivery

ChatGPT Ads are shown inside conversations, and they differ from classic search ads in one fundamental way: the user does not type a keyword, they hold a conversation. The platform decides which conversations your ad fits by reading the context hints you provide.

So approaching it with the Google Ads habit of “let me bid on these ten terms” does not work. A hint is not a keyword list; it is a short description of the moment of need in which your ad makes sense.

Rule one: describe the situation, not the product. Instead of “ergonomic office chair”, write “people who work long hours at a desk and complain of back pain”. Rule two: write exclusions too; saying “not people on a student budget” or “not people looking for second-hand” is the equivalent of a negative keyword on Google.

Rule three: keep each hint to a single moment of need. Cramming five scenarios into one hint leaves the platform unable to match any of them cleanly. Three sharp hints earn more impressions than one blurry one.

Build the structure around moments of need: each campaign is one core need, each ad group a variation of it. For a software product, “people looking for team task tracking” and “people wanting to report on client projects” are two separate campaigns, with different copy and landing pages.

Ad copy should match the conversational tone: short, direct, naming the next step. Loud campaign language feels foreign inside a chat and does not get clicked.

First: ad approval status; a disapproved ad waits silently. Second: hint scope; a hint that is too narrow or too long gets no matches. Third: budget type and daily limit; a limit near zero is exhausted within hours. Fourth: target region and language; the conversation's language must match the ad's.

Fifth: landing page availability; a slow or erroring page can halt delivery. Running these five checks in order resolves most problems before you ever write to support.

The most common mistake on a new channel is postponing conversion tracking. Before the ad goes live, make sure the conversion event is defined on the landing page; otherwise you will never know whether the channel works.

Through Tecrube's Google tools connection you can add the conversion tag via Tag Manager from chat and publish it after a preview; when the same event is used for Google Ads and Microsoft Ads, cross-channel comparison becomes possible. Tecrube manages Google, Meta, Microsoft, TikTok and LinkedIn ads today; for ChatGPT Ads you set up the measurement side with these tools.

The purpose of the first month on a new channel is learning, not sales: which hints earn impressions, which conversation contexts get clicked, which landing page converts. Keep the budget large enough to pay for that learning; changing the daily limit several times in the first week also resets the platform's delivery learning.

Budget decisions between channels can only be made when conversion data is comparable. Pull cost per conversion for Google and Microsoft from Tecrube with one prompt and set it next to the ChatGPT Ads result, and raising or lowering the new channel's share becomes a calculation rather than a guess.

Keep the habit you have on other channels: before a change, note what you changed and why. Tecrube's activity log does this automatically for Google, Meta, Microsoft, TikTok and LinkedIn; keeping the same discipline by hand for ChatGPT Ads makes “what worked?” answerable at month-end.

Do not enable the ad the moment it is built. Have a colleague read the hints: if they cannot answer “in which conversation would this show an ad?”, the hint is blurry. Keep the preview → approve → apply discipline you know from Google by hand here too; on a new channel the first week's budget is a learning budget, not a sales budget.