Albertsons Can't See Its Own ChatGPT Ad Data. They're Still Spending.
Albertsons is spending at least $200,000 on a platform where they cannot see their own campaign data.
That fact, by itself, probably would not be worth writing about. Brands blow experimental budgets all the time. But Albertsons is not some startup throwing money at the latest shiny object. This is a company that just launched incrementality measurement for in-store campaigns using matched market methodology at the store level. Test stores vs. control stores. Actual causal measurement. The kind of rigor most retail media networks are not offering yet.
And they are simultaneously investing in a channel where a technical glitch in OpenAI's Ad Manager is preventing advertisers from viewing their own performance data. One enterprise advertiser in the pilot reportedly spent just 3% of a $250,000 budget over several weeks because they could not calculate ROI.
So why is the most measurement-forward grocery chain in America still spending?
The pilot results, honestly, are rough
Albertsons joined OpenAI's ChatGPT ad pilot back in February alongside Target and Williams-Sonoma, with a $200,000 minimum spend per brand. They started with Valentine's Day. Users typing things like "best flowers for Valentine's Day" or "chocolates and sweets as a gift" into ChatGPT could see ads from Albertsons banners: Safeway, Vons, Jewel-Osco, Shaw's, ACME, Tom Thumb.
The format is contextually triggered. Your ad shows up when someone's conversation signals purchase intent. It is targeting by natural language instead of keywords. On paper, that sounds like an upgrade. And sometimes it is.
But the early numbers are uninspiring. ChatGPT ads are producing a 0.91% click-through rate, according to Adthena research. Google Search averages 6.4%. That is roughly a 7x gap.
Impressions are growing fast though. Volume jumped about 600% from early to mid-March, with Dentsu reporting week-over-week increases. Ads now reach about 5% of ChatGPT mobile users, up from 1% at the beginning of the month.
There is a ceiling problem too. Ads only appear to free-tier and ChatGPT Go users ($8/month). Every Plus, Pro, Business, and Enterprise subscriber is permanently excluded. Which means the reachable audience is, almost by definition, the segment of ChatGPT's user base with less commercial intent. Not exactly where you would want your grocery ads, but that is the deal OpenAI struck to keep premium users happy.
Retail and grocery brands make up about 44% of all advertisers in the pilot. Grocery is overrepresented, which makes sense if you think about it. People genuinely ask ChatGPT for recipes, meal plans, and party ideas. That creates a more natural ad surface than someone debugging code or writing an essay.
Albertsons' measurement obsession is the context that matters
Jennifer Saenz, Albertsons' chief commercial officer, described the pilot as a way to "meet customers when and how they choose to shop for convenient and easy recipes, meals, gifts and party essentials." Standard press release framing. Nothing you would not expect.
But look at who is saying it.
Albertsons Media Collective operates across 2,200+ locations in 35 states and has signed more than 50 advertising partners. They are adding roughly 800 stores to their in-store display network this year. And the incrementality measurement they launched in January is genuinely sophisticated. Matched market methodology at store level: test stores running campaigns compared against control stores with no media exposure. The delta is the incremental lift attributable to advertising.
This is a company that built its media business on the premise that you should be able to prove what works. And right now, with ChatGPT ads, they cannot prove anything. The dashboard is broken.
I think that tension tells you more about where ChatGPT advertising is heading than any CTR metric.
The intent signal might be worth more than the click
My read is that Albertsons is not chasing clicks here. They are chasing something more subtle: the intent data embedded in conversational queries.
When someone types "best flowers for Valentine's Day" into Google, you get a keyword. When someone types the same thing into ChatGPT, you often get a full paragraph. "I want to surprise my wife with flowers, she likes peonies but I am not sure what is in season, and we are hosting a dinner party on Saturday so maybe something that works as a centerpiece too." That is not a keyword. That is a customer brief.
Albertsons can learn more about what their customers actually want from one ChatGPT query than from a hundred keyword bids. Whether they can act on that data is a separate question. And right now, they probably cannot, because (again) the dashboard does not work. But the signal is there, and I suspect that is what keeps the money flowing.
We wrote about the fundamental conflict in OpenAI's ad model a few weeks ago: putting ads inside an "objective" AI assistant creates a trust problem that does not have an obvious solution. The Albertsons experience sharpens that conflict. The retailer is investing because the intent data is rich. But the richer the intent data gets, the more users will question whether ChatGPT's recipe recommendations are organic or influenced by grocery ad dollars.
One more channel to manage, and you cannot even measure this one yet
Most brands running retail media in 2026 are already working across six networks on average, projected to reach 11 by year-end. Amazon measures at SKU level. Meta measures by audience. Google measures by keyword. Now ChatGPT measures by... well, nothing yet.
That fragmentation is familiar to anyone running multi-channel retail media. Each network uses different attribution windows, different definitions of success, and different levels of transparency. Albertsons' own retail media arm is ahead of most on measurement. But even they cannot apply their incrementality framework to ChatGPT spend right now.
Target, the other major retailer in the pilot, has seen ChatGPT traffic to its site grow roughly 40% monthly. That is a real signal, but "traffic" and "incremental sales" are two very different things, and Target has not shared conversion data. When the most successful participant's best metric is "traffic grew," you are still in the brand awareness phase of this experiment.
The $200K question for retail media buyers
The minimum spend keeps this irrelevant for most brands right now. If you are a mid-size DTC brand or an agency managing accounts under seven figures, this is not your fight yet.
If you run retail media for a grocery, CPG, or big-box brand, here is what I would actually do: watch Albertsons' next earnings call for any performance signal. They are the most measurement-forward retailer in the pilot, so if real data surfaces, it will probably come from them. Meanwhile, check your GA4 referral data for any traffic from ChatGPT. Even if you are not in the pilot, users might be mentioning your brand in conversations and clicking through. That referral data is directional at best, but it tells you whether ChatGPT users are already thinking about you.
The bigger question is not whether ChatGPT ads work at a 0.91% CTR. It is whether the conversational intent data they generate has value even when nobody clicks. Albertsons seems to think so. Given their track record with measurement, that is probably not a dumb bet.
But it is worth remembering that even Albertsons cannot tell you if it is working yet.
