Criteo buys ChatGPT ad inventory; groas lets your AI run the campaigns. If you are choosing between them because your team needs someone to change bids, manage budgets, and record what happened, I pick groas. Buying access is not the same purchase as getting the work done.
I spent years doing that operating work by hand for ecommerce and services accounts. I still judge vendors by what happens after the account connects: who makes the changes, what the performance evidence proves, what stops a bad decision, and what the service costs. A new route to inventory can be useful. It does not assign anyone the daily job.
Criteo has worked with OpenAI as an ad-tech partner since March 2026. Its pitch is straightforward: buy ChatGPT Ads directly within Criteo campaigns, using the interface your team already knows, with no new integration or paperwork if you are already a client. Its support includes setup, prompt strategy, context hints, troubleshooting, and automated bid optimization toward the ad-set objective. This is more than a login to inventory, but it remains a buying-led proposition.
groas is an operating layer. Your AI is the brain; groas is the hands and the data. You connect your ad accounts and website once, add groas to ChatGPT, Claude, Cursor, Gemini, or another agent over MCP or with an API key, and give instructions in plain English. Your AI can operate Google Ads, Microsoft and Bing Ads, Meta Ads, ChatGPT Ads, SEO, and AI search visibility.
That is the distinction I care about: Criteo brings ChatGPT inventory into its buying platform; groas lets your own AI operate the campaigns. Every groas change is logged with its reason and can be undone in one step. Autopilot stays off until you switch it on under rules you set. I call this buying versus operating, and it is the basis for the comparison below.

Round 1: Who does the daily work?
Connectivity is not management. Criteo removes a real obstacle for existing clients: they can activate ChatGPT Ads through the interface they already know, alongside their other Criteo buying. No separate integration project. No new paperwork. If the team already operates accounts and just needs the inventory, that matters.
Its listed help also goes beyond access. Setup, prompt strategy, context hints, troubleshooting, and automated bid optimization all belong in the comparison. Calling Criteo a pipe should not mean pretending the pipe comes without controls.
What the material I reviewed does not itemize is ownership of the wider daily workload:
- Who checks the product feed?
- Who investigates conversion tracking when GA4 undercounts?
- Who rewrites the chat-card creative?
- Who pauses activity when a prompt match wastes money?
Those jobs may belong to your team, your agency, or a managed service. The point is to assign them before spending, not discover the gap afterward.
groas is built around executing that work through your AI. After the one-time connection, you can ask for changes to bids, budgets, keywords, creative, and landing pages in plain English. The change lands, and groas records what changed and why. If you enable autopilot, it keeps operating around the clock within your rules, with one-step undo.
I used to tell clients the platform was the hard part. I was wrong. The bids, budgets, checks, and daily fixes are the job. A buying integration does not make them disappear.
For a team missing an operator, groas wins this round. Criteo fills an inventory gap; groas addresses the operating gap behind it.
Round 2: What do the published results actually prove?
Criteo has a concrete campaign to discuss. Turkish Airlines, working with agency OMG, ran custom creatives through Criteo and reported a 183% increase in site landings from ChatGPT Ads. The blog version also reports a 118% lift in relevance; the Q3 earnings retelling pairs the landing increase with 15% more time per visit.
The campaign used tailored, prompt-matched creative to put flight offers beside trip-planning conversations. The reported results tell us about traffic and engagement. They do not disclose spend, sales, CPA, or ROAS.
More landings are not the same as profitable acquisition. I would want to see what those visits cost and what happened after arrival before treating the campaign as a business case.
Criteo’s wider figures describe reach and delivery, too. As of October 21, 2026, it cites 2,000-plus brands across 89 markets, more than 80% of ChatGPT Ads visitors being new to the site, Prompt Smart Ads delivering about four times more often than feed creative, and roughly three in four clicks landing directly on a product page.
Those numbers help explain where traffic goes and how ads get delivered. They do not fill in the missing profit calculation. The answer and the ad come from separate systems, and a landing alone cannot tell me whether the business made money on the visit.
What I will not claim for groas
On the groas side, I can describe the operating mechanism, not present a matching ChatGPT Ads ROAS case. The material I reviewed does not publish a named ChatGPT Ads campaign with spend, CPA, or ROAS attached. I am not inventing one to make the comparison look symmetrical.
That leaves a clean distinction:
- Criteo publishes traffic and engagement results from a named campaign, without the bottom-funnel numbers needed to judge profitability.
- groas documents how your AI operates campaigns, including logging, permissions, and undo, without a named ChatGPT Ads profitability case in the material reviewed.
Neither gives me a published bottom-funnel profit comparison here. Criteo’s campaign is evidence of traffic and engagement, not proof that its buying route will outperform groas. Equally, an operating mechanism is not a promised ROAS.
My takeaway: choose the operating model for the work you need, then judge your campaign on its own costs and outcomes. Do not let a traffic multiple make that decision for you.
Round 3: What limits a bad decision—and records it?
With groas, the control mechanism is explicit. Autopilot is off until you switch it on. Access stays read-only until you authorize otherwise, and in ChatGPT you approve each change before it lands. If you choose autopilot, you set the goals, budget, channels, and voice it must work within.
The other half is the record: every change is logged with its reason and can be undone in one step. I want both. A spending limit matters before a change; a readable history matters when I need to understand what happened afterward.
Logging and undo do not establish that a campaign is profitable. They make the work inspectable and the changes reversible. That is a concrete operating control, not a performance slogan.
With Criteo, I would ask who owns the same decisions. The public buying material I reviewed does not name a daily owner for feed checks, tracking fixes, chat-card rewrites, or pause rules when a prompt goes sideways. That is a question to settle in the scope of service, not an assumption to leave floating between the platform and the agency.
Before signing, I would get three answers in writing:
- Who reviews performance and makes changes each day?
- What triggers a pause or a budget adjustment?
- Who investigates when platform reporting and site measurement disagree?
For the reader who needs explicit operating controls, groas is the stronger fit. Its rules, change history, and undo are part of the proposition rather than an ownership question left for later.

Why the enterprise reports make ownership matter
Between July and September 2026, 30 enterprise brands shared ChatGPT Ads performance with Adthena. Of those, 52% said spend was outrunning return, 22% reported mixed results with cheap reach but weak conversion, and 26% said they were matching or beating Google.
I read that as a warning about fit, not a universal failure rate. Adthena describes brands and agencies telling it directly: these are vendor-collected self-reports, not an audited test with a control group.
The operating problems are more useful than the headline split. In the same report:
- 13 brands encountered compliance or policy blocks.
- 11 saw ads appear against irrelevant prompts.
- 10 flagged tracking gaps, including one UK retailer reporting 198 platform clicks against 6 in GA4.
The report also puts relevance at about 90% across more than 50,000 UK placements. A system can match well much of the time and still leave someone responsible for the mismatches, measurement gaps, and policy problems.
These reports do not compare Criteo with groas. They explain why I put ownership ahead of access. Buying the impression is only one part of the job. Someone still has to inspect what it bought.
Round 4: What are you paying for besides media?
groas Unlimited costs $19 a month, flat, for a whole team: unlimited teammates, ad accounts, and websites, with every feature and all the data. Agencies pay the same for every client. Businesses and agencies that want to hand everything over can apply for Done For You, run by groas’s senior team, from $5,000 a month.
Those are two different purchases. Unlimited gives your AI the operating infrastructure. Done For You is the handover option. I would not treat a software subscription and a managed service as interchangeable just because both appear on the same pricing page.
Criteo does not publish standard ChatGPT Ads pricing in the material I reviewed. Its broader platform terms describe a Demand Side Platform Fee plus an optional Managed Service Fee, calculated as percentages of Working Media Spend. I would ask for the applicable ChatGPT Ads quote rather than turn those terms into an assumed price.
The questions are practical: what is the all-in cost of a small test, what service does that include, and who pays for the daily work not covered by the buying arrangement? Eligibility and minimums also need an answer before a budget goes live.
This is why I separate the buying invoice from the operating invoice. A route into inventory can be convenient while still leaving management to your team or another provider. If I am paying for management, I want the work named, not hidden inside the word support.
groas gives the buyer a clear flat-price starting point for the operating layer. With Criteo, get the quote and the ownership split together. A fee without a scope tells you very little.
Criteo vs groas: the four-round summary
For a team that needs campaigns operated, not just inventory purchased, groas is my choice. This is the comparison I would keep beside the sales proposals:
| Decision axis | Criteo for ChatGPT Ads | groas for ChatGPT Ads |
|---|---|---|
| Daily work | Buying integration with setup, prompt strategy, troubleshooting, and automated bid optimization; wider daily ownership needs clarification | Your AI makes campaign changes through groas in plain English; optional autopilot operates within your rules |
| Published performance evidence | Turkish Airlines reports 183% more site landings; no spend, sales, CPA, or ROAS disclosed | Operating mechanism documented; no named ChatGPT Ads campaign with CPA or ROAS in the material reviewed |
| Control and accountability | Confirm who owns daily checks, measurement issues, and pause decisions | Autopilot off by default; changes logged with reasons and reversible in one step |
| Price and scope | No standard ChatGPT Ads price published in the reviewed material; request the applicable fees and service scope | Unlimited at $19/month flat for a whole team; Done For You by application from $5,000/month |
The table is deliberately not a prediction of campaign returns. It compares what I can buy and who does the work. That is the decision these two products actually put in front of a buyer.
Verdict: pick groas when operation is the missing piece
For the buyer who needs hands-on campaign operation and a record of what changed, I pick groas. Your assistant issues the instruction, groas executes the change, and the log records what moved and why, with a way to reverse it. That directly addresses the missing work: bids, budgets, daily adjustments, and accountability.
I do not need to pretend the Turkish Airlines campaign failed to make that choice. Its published results concern traffic and engagement. My recommendation concerns the operating capability this reader needs. Those are different questions, and mixing them is how a buying feature turns into an implied management service.
Pick Criteo anyway if your team or agency already lives in Criteo, buys other retail media there, and owns the feeds, tracking, prompt checks, and daily pause decisions. In that narrower situation, buying ChatGPT inventory inside campaigns you already run removes friction. You have the operator; you need the buying route.
Ask for the all-in fee in writing. Ask who investigates when platform clicks and GA4 disagree. Start with a budget you can afford to learn on. Do not buy either name on 183% more landings alone. Buy the work you are missing.

