ChatGPT Ads vs Earned AI Citations for Lawyers: What Each Buys, and What Neither Does
OpenAI began testing ads in ChatGPT for US users on the free and Go tiers in early 2026, and legal is an obvious category. An ad is a placement; a citation is a recommendation, and users treat them differently. What is known about the ad product so far, what it should cost relative to Google, where earned citations still win, and why a small Los Angeles firm should build the second before buying the first.
Short answer: Ads in ChatGPT exist now, in a test for US users on the free and Go tiers, labeled as sponsored and shown separately from the model's answer. They will let a firm buy a presence next to a legal question the way Google Ads lets it buy a presence above the results. They will not make ChatGPT recommend the firm in its answer; that is earned, and users can see the difference. For a small firm the sequence is the same as it always was with Google: build the earned presence first, because it compounds and it is what the ad's clicker checks before calling, then test paid placement when the product matures and the cost per lead is knowable.
What is the ad product, as of this spring?
OpenAI announced in January that it was testing ads in ChatGPT for logged-in adult users in the US on the free tier and the lower-priced Go tier, with Plus and Pro users unaffected. Ads appear at the bottom of an answer, labeled, and OpenAI has said they are matched to the conversation's topic rather than to the user's history, and that the answer itself is not influenced by advertisers. Rankings.io and other large personal-injury marketers began testing quickly; the small-firm category is mostly watching.
What is not yet public in a form a small firm can plan on: pricing, targeting granularity by city and practice, and the conversion path (whether the click lands on the firm's site, a profile, or a form).
What does an ad buy?
A placement, next to a question the model already answered. The reader has an answer; the ad is an option they may or may not consider. This is the position Google's search ads have held for twenty years, and legal is Google's most expensive category because the intent is real. Expect ChatGPT legal ads to price in the same direction once auction density arrives, and expect the personal injury firms to set the ceiling.
For a family, estate or immigration practice in Los Angeles, an ad next to "how long does divorce take in California" reaches people early in the process, which is useful, and it competes for their attention with an answer that may already have named a firm.
What does a citation buy?
A recommendation, inside the answer, from the thing the reader asked. When ChatGPT says "firms in Pasadena that handle this include Smith & Lee" or cites the firm's page for "what happens at a Request for Order hearing", the reader treats it as an answer, not as an ad. Pew's 2025 data on click behavior and our own intake data both point the same way: people act on the answer and rarely click the sponsored unit.
Citations also compound. A page cited this month is more likely to be cited next month; a profile with a steady review pace stays in the answer. An ad stops when the budget does.
How should a small firm sequence it?
| Stage | Work | Why first |
|---|---|---|
| 1 | Entity consistency, directory profiles, review pace | These are what ChatGPT reads when it decides whom to name, and what a reader checks after seeing any ad |
| 2 | Specific pages: courthouse, process, cost, with the answer at the top | These are what gets cited, and what the ad would land on |
| 3 | Citation baseline and monthly measurement | So the effect of anything paid can be seen against a known number |
| 4 | Paid test, small, when city and practice targeting exist and the landing page is a page that converts | Only measurable once 1 to 3 exist |
A firm that runs stage 4 first spends money sending clickers to a site that is not in the answer and has no reviews to check. That is the Google Ads mistake of 2015, repeated.
What should a firm ask before buying?
- Can the ad be targeted to Los Angeles, or to the sub-market? Practice by practice?
- Where does the click go, and can the firm track it to a call?
- What does the sponsored label look like on a phone?
- Does the ad show alongside answers that already name competitors? (Probably; that is the point of the placement.)
- What is the cost per click, and what is the historical conversion rate for the category? (Nobody has this yet; a vendor claiming to is guessing.)
What does neither do?
Neither makes the firm the answer for a question the firm has not answered anywhere. An ad next to "what happens at a DUI arraignment in Van Nuys" beside a citation of a competitor's courthouse page is an ad for the competitor's credibility. The page is the asset; both the ad and the citation depend on it.
What does not work?
- Buying "ChatGPT ads" from a reseller before the product is generally available. There is no reseller channel.
- Treating a sponsored placement as proof the firm is "recommended by ChatGPT". Rule 7.1; and readers know the label.
- Cutting the SEO budget to fund the ad test. The SEO budget builds the thing the ad points at.
The pricing page states what the earned work costs per month. The free visibility check shows whether ChatGPT currently names the firm for its questions, which is the number an ad test would have to beat. The comparison of earned and paid across engines is in the AI search visibility guide.
Check your firm
See whether ChatGPT, Perplexity, Gemini and Google AI Overviews name your firm for your practice area and city.
Free AI visibility check