6 Months Into ChatGPT Ads, Advertisers Still Don’t Know What ‘Good’ Looks Like

ChatGPT Ads has changed rather a spot since OpenAI began testing ads successful February.

The level now gives advertisers much ways to buy, optimize, and measurement campaigns, pinch self-service entree disposable crossed 52 countries to date.

We besides person thing we didn’t person astatine launch: existent results from advertisers who person spent money connected the platform.

Some are reporting costs per click (CPCs) beneath $3. Others are person to $10, aliases moreover $13. There are examples of campaigns generating qualified leads astatine competitory costs, on pinch others that produced small meaningful activity aft the click.

The problem is knowing what to make of those differences.

OpenAI says it does not yet person capacity benchmarks crossed advertisers, industries aliases run types. There is besides constricted competitory aliases auction-level reporting to thief explicate why 1 advertiser mightiness salary substantially much than another.

Six months in, advertisers person much ChatGPT Ads information to activity with. But, they still person very small discourse for deciding what “good” capacity should look like.

What Advertisers Can Actually Measure In ChatGPT Ads

ChatGPT Ads Manager gives advertisers astir of the basal capacity metrics needed to measure their ain campaigns.

Reporting is disposable astatine the campaign, advertisement group, and advertisement levels, including:

• Impressions.
• Clicks.
• Spend.
• Click-through rate.
• Average CPC.
• Average costs per mille (CPM).
• Conversions.

Advertisers tin besides export capacity information to CSV for study extracurricular the platform.

Conversion search is disposable done the OpenAI Pixel and Conversions API, pinch support for actions including purchases, leads, and sign-ups. OpenAI whitethorn besides see modeled conversions erstwhile available.

For ecommerce advertisers, reporting tin see attributed sales, costs per attributed waste and ROAS erstwhile those metrics are available.

That’s astir arsenic acold arsenic the existent reporting goes.

There isn’t an Auction Insights equivalent, Impression Share aliases competitory reporting. OpenAI besides hasn’t published capacity benchmarks crossed advertisers, industries aliases run types.

Those gaps make it difficult to put moreover a straightforward metric for illustration CPC into context, particularly erstwhile the auction itself gives advertisers comparatively small visibility into what influenced the value they paid.

See also: How To Measure PPC Performance When AI Controls The Auction

What Advertisers Can’t See Behind The Auction

The deficiency of competitory reporting matters much erstwhile you see really ChatGPT Ads decides which ads to show.

OpenAI uses a relevance-weighted, second-price auction. Ad action is based chiefly connected the discourse and intent of the conversation, on pinch signals from the landing page, advertisement creative, and advertiser-provided discourse hints.

Those discourse hints springiness advertisers a measurement to picture applicable products, needs aliases situations astatine the ad-group level, but they aren’t keywords. OpenAI besides makes clear that they don’t guarantee transportation against circumstantial words, audiences aliases conversations.

That gives advertisers little visibility into the individual matching decisions happening earlier an advertisement ever enters aliases wins an auction.

If CPC rises, for example, the disposable reporting provides small thief separating changes successful title from relevance, inventory, bidding aliases the types of conversations being matched.

The second-price auction adds different furniture of complexity. Advertisers cognize their maximum bid and what they yet paid, but they can’t spot the competitory unit down that value aliases really relevance affected the outcome.

That makes individual CPCs difficult to construe extracurricular the discourse of the advertiser’s ain campaign.

And the results advertisers person published truthful acold show conscionable really wide those differences tin be.

What Early Advertiser Results Show So Far

We person much first-hand ChatGPT Ads information than we did a fewer months ago. Some advertisers person besides spent capable that we’re getting beyond the $500 trial campaigns that were communal early on.

The results still alteration rather a bit, arsenic seen beneath from a fewer advertisers who person shared their data.

Hostinger Spent Nearly $70,000 Testing ChatGPT Ads

Hostinger has 1 of the larger publically shared tests I’ve recovered truthful far.

Hüseyin Ograk, caput of PPC astatine Hostinger, initially reported that CPCs weren’t higher than Google Search and that the campaigns had already generated purchases. He besides described beardown engagement from the traffic.

His appraisal became much nuanced arsenic walk increased.

After astir $70,000 successful testing, Ograk reported CPMs supra $65 but said CTR was a bigger concern. More circumstantial usage cases performed better, while broader messaging consistently struggled.

He besides noted inconsistent traffic quality and challenges evaluating return connected advertisement walk based only connected nonstop conversions.

Similar Tests Are Producing Very Different Results

Other published tests show really difficult it is to tie conclusions from CPC alone.

Common Thread Collective shared results from a precocious mean bid worth (AOV) ecommerce advertiser that scaled from $7 per time to much than $1,000 per time complete astir a month. After $9,620 successful spend, the run averaged a $4.41 CPC, and 0.94% CTR across 136,000 play impressions.

CTC reported betwixt $19,000 and $38,000 successful attributed revenue, depending connected the attribution model, pinch estimated ROAS ranging from 3.3x to 6.8x. The squad relied connected Triple Whale alongside ChatGPT Ads reporting to measure downstream performance.

Floyd Blaikie recovered thing different erstwhile her squad looked beyond the metrics disposable successful Ads Manager.

Her B2B run spent astir $7,000 CAD and averaged a $9.29 CPC, $64.34 CPM, and 0.7% CTR. Her squad past utilized visitant deanonymization to place 146 organizations down 336 paid clicks.

The crazy part? Only 5 matched its perfect customer profile.

Blaikie’s outer study raises a bigger interest than the $9.29 CPC itself. Very fewer of the identifiable visitors matched the advertiser’s ICP, and Ads Manager offered nary measurement to uncover that connected its own.

Synter’s test shows different root of variation: geography.

Across $4,428.84 successful spend, Synter averaged a $9.89 CPC overall. But that number changed considerably by market, from $5.10 successful the U.K. to $10.62 successful the U.S. CPC reached $17.59 successful Australia and $22.89 successful New Zealand, though some markets had considerably little volume.

Across these tests, CPCs alteration wide by advertiser, and market. The disposable reporting still doesn’t supply capable accusation to explicate those differences aliases found a reliable anticipation for what advertisers should pay.

OpenAI’s ain $3 to $5 starting maximum CPC proposal doesn’t needfully lick that problem, either.

The $3 To $5 Recommendation Isn’t A CPC Benchmark

One number has already started showing up often successful discussions astir ChatGPT Ads: $3 to $5 per click.

The fig comes straight from OpenAI, but what it represents is important. OpenAI recommends a starting maximum CPC bid betwixt $3 and $5. Ads Manager whitethorn besides supply bid-strength guidance indicating whether a bid is competitory aliases could limit delivery.

That proposal doesn’t correspond an mean CPC crossed the platform. OpenAI explicitly states that ChatGPT Ads doesn’t yet person capacity benchmarks crossed advertisers, industries aliases run types.

The early advertiser results we’ve already looked astatine reenforce really small the $3 to $5 proposal tells america astir what an advertiser will yet pay.

The consequence is that repeated references to the proposal yet portion distant the connection “recommended.” A $3 to $5 starting bid becomes a $3 to $5 “benchmark,” moreover though OpenAI has published nary specified benchmark.

The consequence is that repeated references to the proposal yet portion distant the connection “recommended.” A $3 to $5 starting bid becomes a $3 to $5 “benchmark,” moreover though OpenAI has published nary specified benchmark.

That framing is already narrowing. In August, OpenAI made Maximize results the default bid strategy for eligible caller advertisement groups, which sets and adjusts bids automatically. OpenAI’s archiving states it does not guarantee transportation against a circumstantial CPA, CPC aliases ROAS target. Advertisers who want a difficult ceiling person to opt backmost into manual bidding.

So the number is simply a starting bid, for the advertisers still mounting one.

The assemblage eligible to spot ChatGPT Ads is not the aforesaid arsenic ChatGPT’s full personification base, though the spread is astir creation much than size. Ads reachand Go users, who dress up the bulk of ChatGPT’s audience, but each higher-priced tier is excluded.

See also: The High CPC Paradox: When Expensive Clicks Are A Sign Of Success

ChatGPT Ads Only Reach Part Of The ChatGPT Audience

Advertisers besides request to beryllium observant erstwhile utilizing ChatGPT’s wide assemblage size to measure the advertizing opportunity.

Ads tin look to users onand Go plans only. Plus, Pro, Business, Enterprise, and Edu accounts stay ad-free, and OpenAI doesn’t show ads to accounts identified arsenic belonging to group nether 18.

That intends the full number of group utilizing ChatGPT isn’t the aforesaid arsenic the assemblage presently eligible to spot ads.

Go is itself a paid subscription, truthful it’d beryllium inaccurate to qualify the ad-supported assemblage arsenic group who don’t salary for ChatGPT. Still, respective of OpenAI’s higher-priced subscription tiers are excluded from the advertizing audience.

That raises an absorbing information for luxury brands, high-ticket products, and different advertisers pinch a narrower customer profile.

There isn’t capable nationalist information astir the demographics aliases purchasing powerfulness of ChatGPT’s ad-eligible assemblage to cognize really intimately it aligns pinch those customers. Broader statistic astir ChatGPT users aren’t needfully typical either, since they see group connected plans wherever ads don’t appear.

That doesn’t mean high-ticket advertisers should enactment away. Common Thread Collective, for example, has already reported affirmative results from a high-AOV ecommerce campaign.

But, it whitethorn springiness those advertisers different logic to beryllium cautious astir applying broader ChatGPT assemblage assumptions to their media plans.

Geography narrows the comparison further. ChatGPT aviator launched successful the U.S. successful February alongside Canada, Australia and New Zealand. OpenAI announced description to the UK, Japan, South Korea, Brazil and Mexico successful May, pinch Brazil and Mexico going unrecorded successful August. Self-service Ads Manager entree reached 52 countries connected August 31, and OpenAI says ads are buyable successful complete 40 markets done its income squad and partners.

Between scheme eligibility, geography, and constricted assemblage data, advertisers are still moving pinch a reasonably incomplete image of who they tin reach.

That makes the advertiser’s ain trial creation and occurrence criteria moreover much important.

How To Approach ChatGPT Ads Right Now

The deficiency of established benchmarks shouldn’t support advertisers from testing ChatGPT Ads, particularly those looking to grow beyond their existing channel mix.

Just spell into that trial knowing location are still limits to what the level tin show you.

Set expectations earlier spending originates based connected what would make the transmission worthy continuing for your business. Someone else’s CPC aliases CTR offers small guidance erstwhile we still cognize truthful small astir the competition, audience, and conversations down those results.

Measurement extracurricular Ads Manager should besides beryllium portion of the scheme from the beginning. Some of the early advertiser tests recovered their astir useful insights elsewhere, including gross attribution, engagement data, and identifying which companies really visited the site.

Early results besides merit much scrutiny than they mightiness connected a mature platform. With less diagnostic signals available, it tin beryllium difficult to abstracted a existent capacity problem from differences successful delivery, matching aliases the conversations disposable during the trial period.

Advertisers whitethorn request much grounds earlier deciding why capacity moved successful 1 guidance aliases another.

ChatGPT Ads will proceed changing, and the information disposable to advertisers should amended pinch it. OpenAI has already said further metrics, reporting views, and insights are planned arsenic Ads Manager develops.

For now, location is nary cosmopolitan number that determines whether a ChatGPT Ads run is performing well.

“Good” has to commencement pinch what the transmission produces for your ain business.

More Resources:

• When Advertising Shifts To Prompts, What Should Advertisers Do?
• The Gap Between AI’s Heartwarming Ads And Its Real Goal: Your Job
• Surviving The Impression Squeeze: How Agentic Commerce Is Changing Google Ads In 2026

Featured Image: Roman Samborskyi/Shutterstock