ChatGPT Ads Review: Advertiser Tests, User Survey, Privacy (2026)

We read every published ChatGPT ads test we could find. Most are too small to prove anything, a few of the numbers do not reconcile, and the click-through gap versus search is about 7x. Sources and unknowns included.

Last verified: 22 October 2026. This review is based on public documents, published advertiser tests and a published user survey. We did not run our own ChatGPT ad campaigns, and nothing here is first-hand testing. No affiliate links.

Key takeaways

Most public "ChatGPT ads reviews" are small tests, and several of the zero-conversion results are too small to say anything. A test of about 40 clicks with no sales would happen roughly 45% of the time even if the true conversion rate were 2%.
Click-through rates sit below search. Adthena measured 0.91% for ChatGPT ads against 6.4% for Google, about seven times lower. Similarweb's May 2026 figure for ChatGPT was 0.68%.
Costs are high for the traffic. Reported cost per click ran from about $4.41 to $13, and cost per thousand impressions from about $64 to $120.
The one large positive result we found has numbers that do not add up on their face (see the ecommerce case below).
Users say they are uneasy, but the survey behind that comes from an agency directory, and its sample skews older.
Targeting and privacy claims by OpenAI and its critics disagree on one point that matters: whether turning off personalization stops ads based on your current chat. It does not.

What OpenAI says about its own ads

OpenAI's January 2026 principles say ads do not influence the answers ChatGPT gives, that conversations stay private from advertisers, and that data is never sold to advertisers. They also say ads will not appear near health, mental health or politics, and will not show in accounts that are, or are predicted to be, under 18. Plus, Pro, Business and Enterprise plans are ad-free. At the time, the plan was a test with logged-in US adults on the Free and Go plans.

By 31 August 2026, OpenAI reported an annualized ad revenue run rate of $1 billion in under 200 days, tens of thousands of advertisers, availability in more than 40 countries, and an Ads Manager open to small businesses since May. It said cost-per-click and outcome bidding make up the majority of campaigns. CNBC reported on 21 October that ads had expanded to 89 countries; we could not load the CNBC page ourselves, so treat that count as reported, not confirmed.

On user trust, OpenAI said its "trust, relevance, experience" measures stayed strong, and gave no numbers. That is a claim we cannot check.

One discrepancy to know about: a secondary outlet put the advertiser count above 100,000 and weekly users at 1.2 billion. OpenAI's own post says tens of thousands and "more than 1 billion". We used OpenAI's figures.

What advertisers report

The table lists every published test we found with figures. These are single accounts, run by different people, in different weeks, on different budgets.

Test	Spend	Clicks / reach	CTR	CPC	CPM	Result
Grow My Ads (Aug 2026)	$1,000 in the title, $1,200 in the scoreboard	92 clicks	1%	about $13	$120	0 conversions
adlibrary	$200	about 5,000 impressions, about 40 clicks	0.8%	about $5	not stated	0 conversions
Unnamed advertiser	$477 over six weeks	370 clicks	not stated	not stated	not stated	0 conversions
B2B campaign (Floyd Blaikie, via Search Engine Journal)	about $7,000 CAD	336 paid clicks	0.7%	$9.29	$64.34	146 organizations identified, 5 matched the ideal customer
Common Thread Collective (ecommerce, high order value)	$9,620	about 136,000 impressions per week	0.94%	$4.41	not stated	$19,000 to $38,000 attributed revenue
Hostinger	nearly $70,000	not stated	not stated	not stated	above $65	mixed; broad messaging struggled, specific use cases did better

Why a zero is not always a verdict

A result of zero conversions only tells you something when there were enough clicks to expect some. If the real conversion rate were 2%, then:

40 clicks with no conversions happens about 45% of the time. The adlibrary test cannot separate "ChatGPT ads don't convert" from bad luck.
92 clicks with none happens about 16% of the time. The Grow My Ads result is weak evidence.
370 clicks with none happens about 1 time in 1,700. That unnamed $477 test says more, because it is unlikely to be luck at a 2% rate. It does not say what happens at a 0.2% rate.

The 2% is an illustration, not a benchmark. OpenAI says it has no cross-industry benchmarks yet, and its $3 to $5 figure is a recommended starting bid, not an average cost.

The numbers that do not reconcile

The Grow My Ads post says $1,000 in the title, summary and body, and $1,200 (rounded) in the scoreboard. The reported 92 clicks at about $13 come to $1,196, which fits the larger figure. Its author also says Google Analytics attribution broke after a tracking-parameter problem, with ChatGPT traffic showing as direct, and the post does not make clear whether conversion tracking was set up.

The Common Thread Collective case has a similar problem. $19,000 to $38,000 in revenue on $9,620 of spend works out to 2.0x to 4.0x, but the reported return on ad spend is 3.3x to 6.8x. The gap might come from a different spend window or attribution setup, and the article does not explain it. We would not quote the 6.8x without the source open next to it.

OpenAI's own headline example, a 3x return over 28 days for one ecommerce advertiser, comes with no spend or attribution method.

The B2B result is the most useful one

Of the 336 paid clicks, visitor identification matched 146 organizations. Five fit the ideal customer profile. That is a targeting-quality finding, and it fits what Hostinger described: inconsistent traffic quality and difficulty judging return from direct conversions alone. It also matches the platform's design. Ads are chosen from the conversation, and context hints are not keywords, so OpenAI does not guarantee delivery against specific words, audiences or conversations.

Reporting is thin. Search Engine Journal's list of missing tools includes anything like Auction Insights, impression share, and competitor reporting. In one early case reported by The Keyword, an enterprise advertiser saw 3% of a $250,000 budget spend with no visibility into why.

How ChatGPT ads compare with search

Measure	ChatGPT ads	Google Search
Average CTR	0.91% (Adthena); 0.68% (Similarweb, May 2026)	6.4% (Adthena)
Reported CPC	$4.41 to about $13 in the tests above	not compared in these sources

The CTR gap is real in both data sets we found, and it is large. Whether a lower click rate is acceptable depends on conversion value, which the public tests mostly cannot show yet. OpenAI also claims a 112% CTR improvement for its newer formats, and agencies have posted "max results" claims. MediaPost reported questions about attribution on both. We could not verify either.

What users report

Clutch, an agency directory, surveyed 747 people (628 regular ChatGPT users) in a report dated 13 October 2026. 55% of respondents were aged 45 or older, so this is not a young-user snapshot, and the survey is a self-report.

52% had seen an AI ad. 55% of them recognized it by its label.
74% were confident they could tell an answer from an ad, while 65% said it is hard to tell. Both numbers appear in the report.
86% think ads influence answers. 66% worry about bias. 75% would trust answers less.
68% find ads unacceptable in medical questions, 59% in mental health, and 53% in financial or legal ones.
50% were willing to switch tools, but only 25% said they would most likely switch.
45% see advertising brands more favorably, and 31% less.

The last two lines matter for advertisers. Stated unease did not turn into a brand penalty in this sample, and the share who would likely leave was a quarter.

Privacy and targeting findings

404 Media tested the ad system and reported that advertisers can target groups by email address, phone number or a mobile advertising ID, and that targeting also uses the prompts people type. Gizmodo's recap says the custom audience upload has a 25,000-record minimum, that OpenAI says uploaded files are deleted within 24 hours, and that how the identifiers are hashed is unclear. We could not read the full 404 Media article, which is paywalled, so the upload details come from Gizmodo.

The personalization opt-out is narrower than it sounds. It excludes ads based on past chats, and Gizmodo quotes the setting as noting that ads may still be based on your current chat.

OpenAI's engineering post, as summarized by DigitBin, describes an orchestrator model that summarizes the conversation before retrieval, filtering, scoring and an auction. The post does not give a list of sensitive topics, does not spell out the rule for minors, and does not say where the settings are. Senator Ed Markey's 22 January 2026 letters to seven AI company chiefs, including OpenAI, asked about exactly these gaps: use of sensitive chat content, minors' data, and how users can tell ads from answers.

What we could not verify

Whether ads near health, mental health and political conversations are blocked in practice. OpenAI states a rule; no independent audit we found tests it.
OpenAI's trust metrics, its 112% CTR claim, and the "max results" agency claims.
The CNBC country count and the secondary advertiser and user figures.
Hostinger's spend breakdown beyond the figures in the Search Engine Journal piece.
Any result outside these accounts. Every test above is one advertiser, and none is peer reviewed.

Frequently asked questions

Do ChatGPT ads work?

The public evidence is mixed and thin. One high order value ecommerce advertiser reported positive returns, though its figures do not reconcile. Several small tests produced no conversions, and only one of those had enough clicks to be informative.

Who sees ads in ChatGPT?

Users on the Free and Go plans. Plus, Pro, Business and Enterprise plans are ad-free, according to OpenAI.

Do ads change ChatGPT's answers?

OpenAI says no. 86% of Clutch's respondents believe they do. We found no independent test that settles it.

Can I turn ads off?

OpenAI says users can turn off personalization and clear ad data. Doing so does not stop ads based on your current chat. An ad-free plan removes ads.

How much do ChatGPT ads cost?

Reported CPC ranged from $4.41 to about $13 and CPM from about $64 to $120 in the tests above. OpenAI recommends starting bids of $3 to $5 per click.

ChatGPT Ads Reviewed: What Advertisers and Users Actually Report