ChatGPT now sells ads, and anyone with $25 a day can buy them.
That is enough to make it worth a serious look. Millions of homeowners describe household problems to ChatGPT every day, and some of those problems have six legs.
Our answer is no, not yet.
Not because the platform cannot reach your service area. It can. You can target ZIP codes, and that surprises most people who assume an AI chat product only sells national reach.
The problem is that once your money is spent, you cannot find out what it bought.
What Are ChatGPT Ads?
ChatGPT ads appear below answers shown to eligible users on ad-supported plans.
The ads are labeled sponsored and kept separate from ChatGPT’s organic response. An advertiser cannot pay to change the answer, influence which companies ChatGPT recommends, or buy a position inside the response itself.
A standard ad may include the advertiser’s name, a short headline, a description, an image or logo, and a link to a landing page.
OpenAI decides when your ad is relevant based on the conversation. Advertisers do not receive the user’s prompt, chat history, or any other private conversation detail.
That protects the user. It also shapes everything that follows in this article.
What It Costs to Get In
The barrier to entry is low. The minimum is $25 per day per campaign in the United States, and OpenAI recommends a maximum bid of $3 to $5 per click.
Real click costs from published advertiser tests run wider than that range. A marketing agency testing its own offer paid $1.89 per click. A B2B services firm paid about $9. Another agency averaged about $13. The one pest control campaign we found anywhere paid $9.
Character limits are tight. Ad titles run 16 to 24 characters recommended and 50 maximum. Ad copy runs 32 to 48 recommended and 100 maximum. That is closer to a Google Ads headline than a Facebook post.
One budget detail matters more than it looks. OpenAI’s documentation states that a campaign-total budget is “a total spending limit, not a pacing control,” and that spend “may accumulate quickly.” A multi-city advertiser we reviewed spent their entire $500 test budget in about an hour.
If you set a lifetime budget and walk away, it may be gone before lunch.
Targeting Is a Description, Not a Keyword List
This is the part that trips up experienced Google Ads managers.
Google Search Ads let you select keywords. A pest control company bids on “termite treatment company,” reviews the search terms that actually matched, and adds negative keywords for traffic that will not produce a sale.
ChatGPT Ads do not work that way.
The primary targeting input is a field called “context hints.” You describe the conversations, topics, problems, or phrases where your service may be relevant. OpenAI’s documentation is explicit that these hints guide matching but are not exact-match targeting rules.
There are no negative keywords. There is no search-term report.
An advertiser running the platform daily described the calibration problem bluntly: no matter what he typed into the context hint field, the platform’s recommendation did not change.
Two other targeting dimensions do exist. Geography is one. Custom audiences built from uploaded email or phone lists are the other, though the list-size thresholds put those out of reach for most single-location operators.
You Can Target ZIP Codes. You Cannot Target a Radius.
This is the most misunderstood part of the platform, so here is the precise state of it.
You can select ZIP codes. You can select cities, states, and designated market areas. We downloaded OpenAI’s own location catalog to confirm the granularity: it contains 41,488 United States postal codes, 210 designated market areas, and 57 states and regions.
So a pest control company absolutely can point ads at the places it serves.
What you cannot do is draw a radius. There is no drive-time tool, no map-drawn boundary, no “20 miles from this address.” To approximate a service area you download the location catalog and hand-assemble a list of ZIP codes.
That is tedious rather than impossible. For one branch it is an afternoon of work. For a company with twelve branches and overlapping territories, it is a project.
Sub-country targeting is also United States only. Canada gets provinces. Every other country where the platform operates is country-level.
Then there is the part that keeps this from being solved. Advertisers who tried to steer delivery geographically report that it did not hold.
One advertiser tried to isolate ads to specific designated market areas and said the campaign kept generating clicks from all over the place. Another set location to the United States and recorded a conversion from France. A third split campaigns by city and found some of those city campaigns never spent any money at all.
Some of those reports may describe an older version of the platform. That is exactly the problem. There is no way to check from inside the account.
The Difference Between Pest Research and Pest Control Intent
Pest control companies need to reach people who can buy service inside a defined geographic area.
General interest in pests is not enough.
Both groups may discuss pest-related topics with ChatGPT.
Only the first group is likely to schedule service.
A Google Ads manager separates those groups using keyword match types, negative keywords, and search-term data. A ChatGPT advertiser writes a description and trusts the system to read the conversation correctly.
Geography does not close that gap. You can put your ad in exactly the right ZIP code and have no idea whether it appeared next to “rat removal company” or “best homemade mouse trap.”
The Platform Can Optimize for Conversions Now
Advertisers can choose from three objectives: CPM for impressions, CPC for clicks, and oCPC, which optimizes delivery toward a conversion event you define.
The conversion objective is a real improvement, and it arrived recently enough that a lot of published commentary still says it does not exist.
A clicks campaign teaches the system to find people who click. A conversion campaign should teach it to find people who complete the action after clicking. Those are not the same people, and for a pest control company the difference is the whole business.
Two cautions apply.
Billing is still per valid click, not per conversion. You carry the risk on traffic quality.
More importantly, conversion optimization is only as good as the conversion signal you feed it. Point an optimizing system at unreliable data and it will confidently chase the wrong thing. Which brings us to the real problem.
The Real Problem Is That You Cannot Audit the Traffic
Here is the finding that decided our answer.
Across the advertiser reports we reviewed, the same pattern shows up independently: the click count OpenAI bills for does not match the number of sessions the advertiser can measure.
(83 in Clarity)
from 14,600 page views
A travel and events business spent about $560 and was billed for 155 clicks. Google Analytics recorded 17 sessions. Microsoft Clarity recorded 83.
A coaching business serving eight cities spent $500. OpenAI reported roughly 140 clicks at about $3.50 each. Not one of those clicks appeared in their Google Analytics or their website analytics. In their words: “We spent the full $500, OpenAI reported clicks, and not one of those clicks appeared in any of our own analytics. Zero leads.”
A B2B services firm spent about $500 over a month. The platform reported 53 clicks. Google Analytics recorded 35 sessions.
An e-commerce advertiser spent about $400 and recorded a funnel that cannot be real: 14,600 landing page views, 236 checkouts started, and zero orders created.
Four different businesses, four different analytics stacks, and the gap runs the same direction every time.
Time on site points the same way. The pest control campaign described below averaged 20 seconds. An Australian advertiser measured 23 seconds from paid ChatGPT traffic while organic ChatGPT referral traffic to the same website averaged two minutes and fifteen seconds.
Same source of visitor. Radically different behavior. The paid placement is the variable.
We are not alleging fraud. Gaps like these usually have dull explanations: bot filtering, link prefetching, redirect loss, consent banners blocking analytics, in-app mobile browsers. Any of those could be the cause.
That is the point. Nobody can tell which, because the platform reports impressions, clicks, spend, click-through rate, average cost per click, average cost per thousand impressions, and conversions. Nothing else. No placement report, no conversation report, no query data.
You are billed for a number you cannot reconcile, with no tooling to investigate the difference.
What Happened When an Agency Ran This for a Pest Control Client
One published test exists in our industry. We went looking specifically, and found exactly one.
An agency ran ChatGPT Ads for a local pest control client and reported the outcome in a paid-media community.
One test is one test. It is not proof of anything by itself, and we are not going to pretend otherwise.
What it does establish is that no pest control company has published a result worth pointing to, and the only result that exists cost $500 and produced no calls anyone could trace.
Some Advertisers Are Getting Results, and Nobody Can Explain Why
It would be dishonest to stop at the failures. There are real successes.
A coaching business in the United Kingdom spent just over £1,500, recorded 398 clicks at £3.81 each, generated 21 opt-ins, booked five sales calls, and closed one contract worth £2,500 up front plus £499 monthly. They had collected £3,375 against £1,500 spent, with most booked calls still ahead of them.
An e-commerce advertiser with a 380-product catalog spent $2,000 over 14 days and recorded 47 conversions at a $42 cost per acquisition, 38% below their Meta campaigns.
An agency that has spent roughly $34,000 across client verticals names home services as one of the categories performing best on the platform, alongside car rental and personal injury law.
And then, in the same discussion thread, an operator in specialized home services said this: “I spent $5k over 3 weeks and had the complete opposite experience as you.”
Read those last two together, because that is the honest state of ChatGPT Ads today.
Two advertisers in the same broad category, one succeeding and one failing, and no reporting available that could explain the difference to either of them.
One more thing worth saying plainly: most of the positive public results come from people who sell ChatGPT Ads services, courses, or tools. That is not a reason to dismiss them. Early adopters usually are the people selling the thing. It is a reason to weigh them carefully.
Where the Optimization Loop Breaks
A paid campaign should follow a simple process, and ChatGPT Ads breaks it between steps two and three.
You may know the campaign produced no booked jobs. You do not know which conversations produced the traffic, which context hint pulled it, or whether the clicks you paid for ever reached your website.
The available actions are to change the hints, change the ad, change the offer, change the landing page, change the bid, or rebuild from scratch. You make those changes without knowing which one caused the original result.
Adding a conversion objective does not fix this. It automates a decision you still cannot inspect.
Multi-Location Companies Have No Account Structure
You can create a separate campaign per market with its own ZIP list and its own budget. That part works, and it is genuinely useful.
What does not exist is agency account structure. There is no equivalent of a Google Ads manager account. Advertisers report that each client has to create their own ad account and then invite the agency in as a manager.
For a company running one branch, this does not matter. For a company running twelve, or the agency running them, it is a real operational cost: separate logins, separate billing, separate verification, and no consolidated view of spend.
Combine that with a budget control that does not pace, and a twelve-market buy becomes a job somebody has to babysit daily.
Paid Subscribers Do Not See the Ads
Ads appear only to users on the Free and Go plans. OpenAI’s documentation confirms that Plus, Pro, Business, Enterprise, and Edu accounts do not see ads.
Go is the $8 tier. So the reachable audience is free users plus the cheapest paid tier.
This does not mean free users cannot become pest control customers. Plenty of homeowners use free software and pay for pest control.
It does shrink an already small pool. A pest control company is working with the homeowners inside its service area who have a pest problem right now and are ready to call somebody. Removing everyone on a higher-tier plan narrows that group before the campaign starts.
What Would Change Our Answer
Eight controls decide whether a local advertising platform is manageable for a pest control company. ChatGPT Ads ships three of them.
The five that are missing are the reason for our answer. Trustworthy click data is the one that matters most and the one that has moved least.
The reporting gap is also the most fixable. OpenAI would not need to expose a single private conversation to close it. Anonymous intent categories, showing that 40% of your delivery landed in do-it-yourself conversations, would be enough to manage a campaign properly.
Our Recommendation
Do not spend a pest control client’s budget on ChatGPT Ads yet.
If targeting were the only obstacle, we would be testing today. ZIP-level targeting exists, per-market budgets exist, and conversion optimization exists. On paper the platform is closer to usable than most people assume.
The obstacle is measurement. A pest control company does not buy clicks. It buys booked inspections and recurring accounts, and it has to know what each one costs. That requires connecting a click to a call, a call to a job, and a job to revenue.
The first link in that chain is the one that is broken. Four advertisers independently found the clicks they were billed for did not appear in their own analytics. Nobody can currently explain the gap, and the platform provides no reporting that would let them try.
If you are an owner who wants to try this with your own money, it is a defensible experiment at $25 a day. Build a tight ZIP list. Run a dedicated call tracking number so no lead gets miscredited. Tag every URL. Then reconcile the platform’s click count against your own analytics before you judge anything else. Treat the first few hundred dollars as the cost of learning what the platform tells you and what it hides.
Do not treat it as a lead channel yet, and do not let anyone bill you for managing it as one.
We Are Running Our Own Test
Every number in this article came from someone else’s ad account. That is a weakness, and we are fixing it.
We are running our own ChatGPT Ads test on Pesty’s money, not a client’s. We are not testing whether it converts. A few hundred dollars at these click costs cannot answer that question honestly.
We are testing two things that a few hundred dollars can answer.
Does ZIP code targeting actually enforce? We will build a tight ZIP list around a single market and reconcile where the clicks came from against our own analytics and call tracking.
Do the platform’s reported clicks match measured sessions? Four advertisers say no. None of them were in home services, and none had call tracking in place.
We will publish the numbers here either way, including the ones that make our position look wrong.
If ChatGPT Ads becomes a real acquisition channel for pest control, we want to be early. We just are not going to find out with your money.
