The Shiny Object Problem in Media Buying: How We Decide Which Channels Are Worth Testing

At the end of August, OpenAI announced that ChatGPT Ads had crossed a $1 billion annualized revenue run rate in under 200 days, ads now live in more than 40 countries and self-serve buying is open to advertisers across India, Europe, the Middle East and North Africa. Numbers like that create pressure fast. If a channel is scaling that quickly, shouldn’t every advertiser already be testing it?

Look at the number a second time before you answer. An annualized run rate is the current month multiplied by twelve, roughly $83 million a month, not a billion dollars collected. And Emarketer analyst Nate Elliott told Reuters that the figure indicates OpenAI is tracking behind its own internal target of $2.5 billion in advertising revenue for the year. A real business was built quickly, but by the platform’s own plan, it also came in short.

That gap between the headline and the math is the whole problem in miniature. Every year brings a fresh wave of platforms, ad formats, partners and AI-powered tools promising to change how brands reach customers. It is tempting to treat every new option as an immediate opportunity, and easy to judge an agency by how many channels it is willing to add to a media plan. At Location3, we don’t see it that way. Every channel added to a mix brings budget fragmentation, measurement complexity, creative demands and operational overhead. Our job isn’t to place a brand everywhere it could possibly appear. It is to invest where the evidence says it can move the business, measure what happens, and use that data to get better.

Availability is not the bar

A channel doesn’t earn our recommendation because it’s new, because it’s in beta, or because a competitor started using it last month. Before we bring a test to a client, we want a credible reason to believe it can perform for that specific business. When that case can’t be made, “not yet” is the right answer, and it protects a client’s investment more than a premature test would. When the evidence is there, we move, with a defined plan, clear ownership and a real way to measure whether it worked.

What the ChatGPT Ads case actually looks like

For its first several months, ChatGPT Ads sold on country-level targeting only. All 50 states or nothing, which ruled it out for the vast majority of local businesses and multi-location advertisers. That changed in late May, when OpenAI added targeting by state, DMA and ZIP code. It was a real fix for exactly the kind of client we work with. Worth knowing: the finer targeting is US-only, so a brand with Canadian units is still buying those at the country level. Targeting was never the harder problem. Measurement is.

You can track conversions in ChatGPT Ads. OpenAI ships a pixel and a Conversions API, and Ads Manager reports impressions, clicks, spend, CTR, average CPC, average CPM and one rolled-up conversions number at the campaign, ad group and ad level. It reports nothing beneath that. No query-level data, no placement data, no demographic breakouts. That is a privacy-by-design boundary, not a beta gap that closes next quarter.

Comparing those results against the rest of a media mix is where it comes apart. Clicks out of AI interfaces frequently arrive at a site with no referrer, so GA4 logs the visit as direct traffic. Tag the destination URLs correctly and the sessions still land in Paid Other, because ChatGPT is not on Google’s source list for paid search, paid social or paid shopping. OpenAI’s own conversion documentation acknowledges that Ads Manager, GA4 and other ad platforms will report different conversion totals, because each applies a different attribution methodology. Reporting connectors have started to pick up the channel’s spend and click data, but conversion data is not yet exposed through OpenAI’s API to them.

Add that up and cross-channel comparison is not impossible. It is a custom build. A franchise system can commit three months and real dollars, pay for the measurement work to stitch it together, and still finish with numbers that won’t reconcile against paid search, paid social or local display. That reconciliation is the comparison that tells you whether the budget was well spent. Until it gets easier to produce and easier to trust, we’re not making that bet on a client’s behalf. We are watching it closely. OpenAI has said its next phase of growth brings additional formats, objectives, buying options and measurement capabilities. When the measurement piece lands, we will say so.

How we determine what earns a media recommendation

Every channel is different, and so is every brand. A few questions tend to separate the opportunities worth pursuing from the ones worth waiting on. The first is simple. Is there a real reason to believe this can move a defined business outcome, something beyond novelty, press coverage or a vendor’s own claims? From there, fit matters. A channel built for top-of-funnel awareness isn’t going to solve a near-term conversion problem, no matter how impressive its reach numbers look. And reach itself needs scrutiny. A platform can carry plenty of volume and still lack the audience or the targeting precision a specific business model requires to drive results efficiently. Then there is the question we come back to more than any other. Can we measure the outcome? Can we track results, get transparent data for analysis and reporting, and optimize on signals we trust? If the answer is no, nothing else on the list matters. We won’t recommend client funding for a channel we can’t measure with enough confidence to make a real decision. The ChatGPT Ads case above is that problem playing out in real time.

Those questions aren’t ad hoc on our end. Every emerging channel we evaluate gets scored the same way, against a set of hard-gate disqualifiers and a set of weighted dimensions. The hard gates work the way they sound. A channel that fails one is out, regardless of how well it scores everywhere else. The weighted dimensions are where judgment does the work, and the heaviest weight in our model sits on local and sub-national viability, because most of what we run is franchise and multi-location. A platform that can only be bought nationally starts at a structural disadvantage with us and reach numbers don’t make that up. We keep the scoring itself internal, because the output is a recommendation for a specific brand, budget and objective, not a general verdict on a platform. The questions underneath it travel. If you are a multi-location brand trying to sort a genuine opportunity from a distraction, start here:

• Do you have enough incremental investment to produce a meaningful sample of useful campaign data, without quietly pulling dollars from channels that are already working?
• Can you define a clear hypothesis, a pilot duration, and the conditions under which you would call it a win or walk away?
• Are the platform and the partner stable enough to trust for the length of a test?
• If the pilot succeeds, is there an honest path to repeating it across more locations, or is it a one-off that can’t scale?

None of these questions has a universal answer, and that’s the point. A channel that clears the bar for a specialty retailer or an eCommerce platform can fail it completely for a 400-unit franchise system, and the reverse is just as true. What doesn’t change is the standard every opportunity has to meet before we put media budget behind it, including in a test.

Our agency POV

We evaluate every meaningful new channel, and we say so honestly when one isn’t ready, even when the headlines suggest otherwise. Staying ahead of the market and staying disciplined tend to require each other. We move quickly when a client, an objective, a budget and a way to measure the result all line up. We hold back just as quickly when they don’t. What that protects is straightforward: client media budgets stay directed at what moves the business, instead of wherever the industry happens to be looking this month.