ChatGPT ads can now target your Charleston ZIP

I downloaded a spreadsheet off OpenAI's developer site tonight. It is the list of every place an advertiser is allowed to point a ChatGPT ad at. 42,110 rows.

Row 9029464 is Mount Pleasant, South Carolina.

Folly Beach is in there. So is Johns Island, Daniel Island, Isle of Palms, Sullivan's Island, Goose Creek, Hanahan, Summerville, West Ashley, James Island, and downtown 29401. The Charleston, SC television market has its own ID, 3000019, sitting alongside 209 other metro areas.

That is the news. Not "AI is coming for local search" in the abstract. An actual file, published by OpenAI, with your service area itemized in it.

What is really in the file

Here is the breakdown of those 42,110 targetable locations: 41,488 postal codes, 249 countries, 210 TV regions (Nielsen designated market areas), and 163 regions, which is their word for states and provinces.

One honest caveat, because I would rather be right than exciting. The campaign targeting documentation says OpenAI Ads supports country, region, and DMA targeting. It does not promise ZIP code delivery. But the location catalog they publish from that same page is 98 percent postal codes. Somebody built the plumbing for street-level local targeting and shipped the catalog before the docs caught up.

(I keep the CSV bookmarked now. When those rows become selectable in the interface, that is a real date for local advertising, and I would rather notice it in a file than read about it three weeks later.)

Why this is moving right now

ChatGPT Ads has spent July turning into an actual ad platform. On July 24, Search Engine Land's Anu Adegbola reported eight new advertiser capabilities: a conversion-optimized bidding objective, average daily budgets over a rolling seven-day window, automatic budget pacing, geographic exclusions, AppsFlyer and Adjust integrations, advanced conversion matching, a bulk API, and refreshed product cards. Four days later Barry Schwartz flagged on X that OpenAI had started emailing advertisers directly about the budget change.

Average daily budgets. Budget pacing. Geo exclusions. If those words sound familiar, it is because that is the Google Ads feature set, rebuilt. This is not an experiment anymore.

What a ChatGPT ad actually looks like

Straight from OpenAI's own help documentation, because this is where most of the hype articles get sloppy.

An ad sits below a ChatGPT response and carries six things: advertiser name, favicon, headline, description, landing page, and an image. It is closer to a sponsored result than a banner.

Nobody on Plus, Pro, or any Business plan ever sees one. Neither does any user OpenAI believes is under 18. So the audience is the free tier.

You buy on impressions (the Reach objective) or clicks (the Clicks objective), through a relevance-weighted second-price auction, and OpenAI's stated recommendation is to open with a maximum bid of three to five dollars per click.

And here is the part that should make every contractor pause. There are no keywords. At the ad group level you provide what OpenAI calls context hints, describing conversations and topics where your service might be relevant. Their documentation says plainly that these are not exact-match keywords and do not guarantee delivery in any specific conversation.

Read that again. You cannot buy "emergency plumber Mount Pleasant." You can suggest a vibe and hope the matching system agrees with you.

"But ChatGPT users are ready to buy"

Yes, a lot of them are. That is the strongest argument for this platform and I am not going to pretend otherwise. Someone typing "my water heater is leaking and I do not know if I should shut off the main" is further along than someone idly scrolling.

But run the math like an owner instead of like a marketer. You are paying three to five dollars a click, into an ad-matching system you cannot steer with keywords, for an audience that excludes every paying ChatGPT subscriber, with attribution tooling that is weeks old. Compare that to a Local Services Ad, where you pay per lead, the lead is already geographically qualified, and Google is folding the whole LSA system into Google Ads next month anyway. One of those channels is a known quantity in the Lowcountry. The other is a beta.

If you are working with a real budget instead of an unlimited one, and most Charleston service businesses are, the beta does not get to eat the known quantity's lunch money.

What I would actually do this week

Do not move a dollar out of what is working. Not yet. The targeting file is a signal about where this goes, not an invitation to reallocate in July.

Get citable before you get advertised. The free-tier user seeing ads is the same user asking ChatGPT who to call. Being the business it names organically costs nothing per click and does not expire when you pause a campaign. That is the whole argument behind why Charleston businesses winning the map pack can still be invisible in AI answers, and why the third-party "best of" lists AI models read are worth more of your attention right now than an ad account.

If you want to test anyway, test like an adult. Cap it at 500 dollars. One ad group. Put a dedicated UTM on the landing page URL, because OpenAI confirms tracking parameters persist through the click and will show up in your analytics. Then judge it on booked jobs, not clicks. A roofer who cannot tie a ChatGPT click to a signed contract has bought a lottery ticket, not a lead source.

Bookmark the location catalog. When Mount Pleasant becomes selectable in the interface and not just present in the file, the arithmetic changes fast, and the people who noticed first will get the cheap clicks.

We watch this stuff at Charlestowne Marketing because our clients are the ones who eat the difference when a new channel turns real. If you want the Charleston digital marketing side of your business built to catch that shift instead of chase it, that is the work.

Your action item this week is a small one. Open ChatGPT, ask it who the best company in your trade in Charleston is, and read the answer carefully. If your name is not in it, no ad budget in the world fixes that. That is the job to do first.