Ads Where People Are Thinking: ChatGPT Ads Reach Sri Lanka

Imagine someone typing this into ChatGPT:

“I have about LKR 9 million. Would a hybrid or a petrol car make more sense if I mostly drive in Colombo traffic and visit family in Kandy on weekends?”

That isn’t a search query. It’s a budget, a daily routine and a doubt, all in one message. A search engine would boil it down to a few keywords. A conversation keeps the whole situation.

Now OpenAI has listed Sri Lanka as a country where its self-service Ads Manager is available, which means a labelled ad could one day sit underneath an answer like that one. So marketers are asking the question they ask about every new channel: is it any good?

Here’s my honest answer: it’s too early to know. Let me show you what we do know, what we don’t, and how I’d think about testing it.

A note on evidence: the facts below were checked against OpenAI’s help pages on 23 October 2026. Anything marked as my read is interpretation, not OpenAI’s stated reasoning.

What has actually changed?

OpenAI’s Ads Manager availability page now shows Sri Lanka as “Available” for self-service. Three details matter for anyone planning to try it:

• Ads appear as clearly labelled paid placements below a response, and OpenAI says they don’t influence the answer.
• Advertisers see aggregated performance data, never users’ conversations.
• Ads are shown to people on the Free and Go plans. Plus, Pro and Enterprise subscribers don’t see them.

There’s also a quiet condition: the legal entity that advertises and gets billed has to be based in a listed country. Availability on paper doesn’t mean every business can switch it on.

Why is this different from a search ad?

Think of two short contrasts.

Search: what do you want? Someone types “hybrid car price Sri Lanka” and an ad answers that request.

Conversation: what are you trying to work out? Someone explains their budget, their commute and their worries, and an ad can show up while they’re still deciding.

That’s the appeal: reaching people in the consideration stage, with richer context than a keyword gives you. Whether that context turns into better results for a Sri Lankan advertiser is a separate question, and nobody can answer it yet.

Is this a good time to start?

It’s a good time to learn. It’s not necessarily a good time to scale.

There are no local benchmarks, no Sri Lankan case studies and no clear picture of how many people here will actually see these ads. Anyone declaring it a winner or a flop right now is guessing. A small, well-measured test is the sensible move.

What are the upsides and the unknowns?

What looks promising

• It reaches people while they’re comparing options, with more context than a short query.
• It can reach people who haven’t searched for your brand or category yet.
• There are clear objectives (views, clicks, conversions) and measurement options that are still growing.
• Competition is likely low, so early tests may teach you more cheaply than later ones. That’s a guess, not a guarantee.

What we can’t yet see

• Cost per result, because there’s nothing local to compare against.
• Audience size and language reach in Sri Lanka.
• How many relevant conversations will actually show your ad. There’s no guarantee it appears in every one.
• How mature the support around it is: fewer agencies, fewer case studies, fewer best practices.

Why might the budget feel higher than you’d expect?

OpenAI’s documentation lists a minimum daily budget of USD 25 per campaign for US accounts. The minimum varies by market and billing currency, and I couldn’t find a published LKR figure yet, so check what Ads Manager shows for your own account.

It also helps to know that the daily budget is an average. OpenAI says daily spend can reach up to twice your daily budget, and up to seven times it across a week. So at USD 25 a day, a 30-day test averages around USD 750. For a small Sri Lankan business running one experiment, that’s a real commitment.

Why do entry costs tend to run high on a new channel? This is my read, not OpenAI’s:

• Limited inventory. Ads don’t appear in every conversation, so supply is tight.
• Fixed minimums in foreign currency. They simply feel bigger in local terms.
• Early premium. New formats often start pricier because brands want to be first.
• Learning cost. With no benchmarks, you need enough spend to get data worth reading.

How does it sit next to Google and Meta?

So what should we ask instead?

I’d swap “Are ChatGPT Ads good?” for three questions that can actually be answered:

1. What would success look like for my business?
2. What’s the smallest test that could tell me?
3. Is my tracking ready to read the result?

That’s the same chain I use in the lab: question, hypothesis, experiment, data, result, learning. A new channel doesn’t change the method. It just makes the first question harder, because we’re starting without a baseline.

So here’s the one I’m still turning over: if you could spend only a small, fixed budget on this, what result would change your mind about it, in either direction?

Reference Article: https://help.openai.com/en/articles/20001210-creating-campaigns-for-chatgpt-ads