

ChatGPT ads are sponsored listings that OpenAI shows below ChatGPT's answers to logged-in adults on the Free and Go plans. Since August 31, 2026, US law firms can run them, provided the advertiser is licensed to practice law where the ad is shown and passes OpenAI's manual review. The ad never appears inside the answer itself.
People already bring their legal problems to ChatGPT: a firing after medical leave, a custody schedule that no longer works, a landlord who keeps the security deposit. This guide covers what OpenAI has actually launched, where ads appear, who sees them, how an ad gets picked, how the channel compares with Google Ads, what advertisers can and cannot see, and how to build a first campaign within ABA Model Rules 7.1 to 7.3.
ChatGPT ads are not banners and not sponsored mentions written into a response. They are separate, labeled units matched to the context of a conversation. OpenAI turned them from a test into a full ad platform in less than a year.
| Date (2026) | What OpenAI announced or launched |
|---|---|
| January 16 | Plan to test ads for US users on the Free plan and on Go, its $8-a-month tier |
| February 9 | Ads testing starts in the US, below answers, labeled as sponsored |
| May | Self-serve Ads Manager in beta for US advertisers, cost-per-click bidding, pixel and Conversions API |
| August 31 | Ad policy update permitting legal services in the US; ads reach a $1 billion annualized revenue run rate |
| October 21 | Ads in 89 countries; more than 100,000 advertisers have used the platform |
| November 9 | Target CPC bidding for new click campaigns |
According to OpenAI, ads reached that $1 billion run rate in under 200 days. This is no side experiment: the format is here to stay.
When ads launched, legal services sat on OpenAI's list of categories not allowed. Version 1.5 of OpenAI's ad policies, dated August 2026, changed that. Ads for legal advice, representation, or legal services are now permitted in the US only when the advertiser is licensed to practice law in the jurisdiction where the ad is shown. The policy names immigration, personal injury, legal claims, and document preparation.
Two limits apply. Legal services ads outside the US remain prohibited. And OpenAI rolls out regulated categories gradually, with approvals reviewed manually on a case-by-case basis. Being eligible does not mean being approved on day one.
The core difference with Google: on Google, you buy a keyword, a fragment of a query. In ChatGPT, your ad is matched to a whole conversation in which the person has often described their situation, their doubts, and their constraints. You answer an intent with context, not a three-word search.
Paying does not change what ChatGPT says. OpenAI states that ads do not influence ChatGPT's answers and that advertisers cannot shape or rank responses. Being named in the answer itself is an organic question, covered in our guide to getting cited by ChatGPT. The ad is the paid layer next to it.
Timing is the other reason to look now. The legal category opened less than three months ago, and OpenAI publishes no benchmarks across advertisers, so nobody can say how crowded legal auctions are. Firms that learn the format early will have their own data when competition arrives.
Before you spend a dollar, know exactly where your firm will show up. The answer is simpler than many agency pitches suggest: one placement in conversations, with a fixed format.
According to OpenAI's help center, ads appear below ChatGPT's response, labeled as sponsored and visually separated from the answer. Each ad shows the advertiser's name and favicon, a title, a short description, an image, and a link to a landing page. Users can dismiss an ad, see why it was shown, and send feedback. OpenAI has also announced a visual format shown during image generation, which has little use for a law firm.
For a prospective client, the ad reads as a next step after an answer, not as an interruption.
| Element | What OpenAI specifies | What it means for a law firm |
|---|---|---|
| Placement | Below the response, labeled as sponsored, never inside the answer | Write the ad as a next step, not as legal advice |
| Advertiser name and favicon | Shown on every ad | Use the firm's real name, which also serves Rule 7.2's name requirement |
| Title | 50 characters maximum, 16 to 24 recommended | Name the situation, not the firm's slogan |
| Description | 100 characters maximum, 32 to 48 recommended | One fact and one action, no promise of results |
| Image | One square image, or 4 to 6 square images in the multi-image format | A real photo of your lawyers or office beats a stock gavel |
| Landing page | Must be reachable by OpenAI's review crawler and comply with its ad policies | One page per situation, with your contact details visible |
No ad is written into the body of the answer, no label says "recommended by ChatGPT," and no paid option ranks firms inside a response. Anyone selling placement inside ChatGPT's answer is selling something OpenAI does not offer. Your ad also cannot claim or imply that ChatGPT recommends your firm: under Rule 7.1, that would be a misleading communication.
ChatGPT's reach is huge, but the audience your ad can reach is narrower than the headline number. Here is who is in, who is out, and what a conversation tells you.
OpenAI says more than 1.2 billion people use ChatGPT each week (October 2026). Ads reach only part of them: logged-in adults on the Free and Go plans. Plus, Pro, Business, Enterprise, and Edu accounts see no ads, and neither do accounts identified as belonging to users under 18. In other words, the premium subscriber who pays for ChatGPT Plus is exactly the person you cannot reach with an ad.
OpenAI also keeps ads away from sensitive topics, naming health, mental health, and politics. In June 2026, before the August policy update, an OpenAI spokesperson told Marketing Brew that the company was not widely placing ads in legal, financial, or medical conversations, only in limited cases where they fit the context. Expect modest volume at first.
Pew Research Center surveyed 5,119 US adults from February 17 to 23, 2026, and found in June 2026 that 44% had used ChatGPT. Use falls with age: 61% of adults 18 to 29, 55% of those 30 to 49, 37% of those 50 to 64, and 19% of those 65 and older.
That profile fits practice areas whose clients skew younger: employment, immigration, landlord-tenant, family law, and consumer disputes. It fits less well for estate planning aimed at retirees, where Google search and referrals still dominate.
A typical conversation where a family law ad could be relevant:
"My husband and I want an uncontested divorce in Ohio, but we have two kids under five and a house we bought together three years ago. What are the steps, how long does it take, and do we each need our own lawyer?"
This one message reveals a legal step that is coming soon, a state, a property issue, children, and a wish to understand before acting. No Google keyword carries that much context in a single query.
Someone who asks three questions about a wrongful termination, then "how do I find a good employment lawyer in Chicago," is ready to decide. Whether an ad shows at that moment depends on OpenAI's eligibility rules for the conversation, not on your bid alone. The audience is smaller than Google's, but it arrives with its problem already put into words.
OpenAI built its own ad stack around a post-trained language model. In OpenAI's description, it finds candidate ads, predicts engagement, judges how useful each ad is for the conversation, filters for relevance and safety, then runs a relevance-weighted, second-price auction with a minimum relevance-and-value threshold. OpenAI does not publish the weight of each factor, so treat any precise percentages you read elsewhere with suspicion.
| Input | What OpenAI says | What to do as a law firm |
|---|---|---|
| Conversation context and intent | Ads are matched to the topic and intent of the current conversation | Build each ad group around a situation clients describe, not a practice area label |
| Context hints | Plain-language descriptions of relevant conversations; not exact-match keywords and no guarantee of delivery | Write one clear phrase per idea, in the words a client would use |
| Ad copy and landing page | Both feed ad selection and policy review | Make the ad and the page answer the same situation |
| Bid | The auction sets the price; a high bid does not rescue an irrelevant ad | Fix relevance first, then adjust bids |
| Personalization | Signals from past chats are used only when the user keeps ad personalization on | Write for the conversation in front of you, not for a profile |
On Google, a family law account is often split by service: divorce, custody, child support. In ChatGPT, the efficient split is the situation the future client lives through: a high-conflict separation with young children, a dispute over the house, a parent who wants to relocate with the kids. Each situation calls for its own ad group, its own ad, and its own landing page.
OpenAI's help center supports this logic. It recommends a separate ad group when hints cover meaningfully different needs or audiences, and a campaign can hold up to 10,000 ad groups. It also notes that people do not need to use your exact words for their situation to match.
Hints do not enforce location, schedule, or exclusions. Geographic, platform, and audience targeting are set separately in the campaign, which matters for lawyers: OpenAI's licensing rule follows the jurisdiction where the ad is shown, so target only the states where your lawyers are admitted. And keyword stuffing helps nothing: OpenAI asks for a natural phrase focused on one idea, not a list of disconnected terms.
Three channels, three logics. On Google, you buy a moment of search: you know what the person typed, nothing more. On Meta, you target a profile and interrupt someone scrolling a feed. In ChatGPT, your ad appears after a dialogue in which the person has already described a problem and asked follow-up questions.
| Criterion | Google Ads | Meta Ads | ChatGPT ads |
|---|---|---|---|
| Targeting unit | Keywords and match types | Audiences built from interests, behavior, and your own data | The conversation's context, guided by your context hints |
| User mindset | Active search, sometimes quick and vague | Passive scrolling and discovery | Working through a problem, step by step |
| Where the ad sits | Above and below search results, on Maps | In the feed, Stories, and Reels | Below the answer, labeled as sponsored |
| How you pay | Mostly per click | Usually per impression, with click and conversion objectives | Per 1,000 views, per click, or per click optimized for conversions |
| Access for law firms | Open; Local Services Ads add Google screening | Open, subject to ad review | US only, licensed in the jurisdiction, manual review |
| Public cost benchmarks | Yes, and legal clicks rank among the most expensive | Yes, by industry | None yet, according to OpenAI |
Google has the volume, years of legal benchmarks, call tracking, and Local Services Ads that charge per lead. It is also the most expensive auction in the market: attorneys and legal services have the highest average cost per click of any industry in LocaliQ's 2026 benchmarks, as our breakdown of Google Ads cost for lawyers shows. For most firms, Google remains the main paid channel.
Context. The person has often written several sentences about their situation before your ad appears, which should mean fewer clicks from people looking for something else. The auction is also young. Whether that translates into a lower cost per signed case is not proven yet: no public data exists for legal campaigns, so you will have to measure it on your own account.
Treat ChatGPT ads as a test next to Google, not a replacement. The likely pattern is fewer contacts than a Google campaign at the same budget, from people who have already described their need. Compare both channels on the same situations, with the same definition of a qualified lead, over at least two to three months.
Confidentiality is a fair question for a profession bound by Rule 1.6. Here is what OpenAI says it does, and where your own duties begin.
OpenAI states that it keeps conversations private from advertisers and never sells user data to them. Advertisers get aggregate reporting only: impressions, clicks, spend, click-through rate, average CPC and CPM, and conversions. On the user side, people can clear their ad history, turn off ad personalization, and see why a given ad appeared.
What your firm sees, and what it doesn't
As an advertiser, you never see what a user wrote. You see volumes: how many people saw the ad, how many clicked, what it cost, and how many conversions your tracking recorded. The mechanism is close to Google Ads, where you see aggregated search data, not a person's history.
Once a visitor lands on your page, privacy is your job. OpenAI offers a pixel and a Conversions API to report conversions. Rule 1.6 protects all information relating to a representation, and state laws such as the CCPA in California govern the personal data your site collects. Send events, such as a form submitted or a call started, never what the prospect wrote about their matter. The same rules as for Google Ads AI tracking apply here.
Buying ads does not put any firm or client data into ChatGPT. Using ChatGPT for client work is a separate question, with its own ethics rules and plan choices, covered in our guide to ChatGPT for lawyers. Note that the Business and Enterprise plans many firms use are ad-free, so your own team will not see the ads you run.
US advertisers can register in Ads Manager, OpenAI's self-serve platform, still labeled beta, or buy through OpenAI's partners. For a law firm, expect manual review and be ready to show that your lawyers are licensed in every state your ads will reach. Here is how we would structure a first campaign.
An effective ad combines four elements: the situation or practice area, named plainly; the place you serve; one verifiable fact, such as video consultations or appointments within the week; and a direct action, such as booking a consultation. Avoid wording that could fit any provider.
Then check the rules. A ChatGPT ad is advertising under Rule 7.2: it must be truthful and not misleading under Rule 7.1, and include the name and contact information of a lawyer or firm responsible for it. ABA comment [12] to Rule 7.2 accepts a website address as contact information. No guarantees, no unverifiable "best," and no "specialist" claim unless you hold a recognized certification. Our guide to lawyer advertising rules covers the details.
One setting matters more than any other. On October 30, 2026, OpenAI turned on creative enhancements by default for self-serve campaigns: they can rewrite your headlines and descriptions, and the new versions can run without your approval. Turn them off under Manage settings in Ads Manager, because you answer for every word under Rule 7.1, whoever wrote it.
Two state-level points remain. Some states require filing: Florida asks for most ads at least 20 days before first use, Texas within 10 days after. On solicitation, comment [1] to ABA Model Rule 7.3 excludes communications automatically generated in response to electronic searches, and a ChatGPT ad follows a conversation the user started. Check how your state's version reads.
| Element | Recommendation | Example: employment law firm in Chicago |
|---|---|---|
| Ad group | One situation per ad group | Fired after returning from medical leave |
| Context hint | One natural phrase, one idea | "An employee in Illinois who was fired after coming back from medical leave and wants to know if that was legal" |
| Location | Only states where your lawyers are licensed, and only where you take cases | Illinois, narrowed to the Chicago area if you meet clients in person |
| Bidding | Pay per click, with a target CPC | Start near the $3 to $5 per click OpenAI's help center cites as a starting bid, then reset the target to your real average after a few weeks |
| Ad copy | Situation + place + fact + action, within 50 and 100 characters | Title: "Fired After Medical Leave?" Description: "Chicago employment law firm. Talk to an attorney about your case this week." |
| Landing page | A page for the situation, never the homepage | "Fired after medical leave in Illinois," with a short form and a visible phone number |
| Metrics | Click-through rate, cost per lead, cost per consultation, cost per signed case | Goal: a cost per consultation below Google Ads on the same situation |
On bidding, OpenAI made target CPC available for new click campaigns on November 9, 2026, and plans to retire maximum CPC bids on existing campaigns on January 1, 2027. A target CPC is an average, not a cap: single clicks may cost more or less, and a target set below your real costs can restrict delivery. Billing is postpaid, charged as your spend crosses a payment threshold.
Sending a prospect to your homepage after a context-rich ad wastes most of that context. The page should continue exactly the situation that triggered the ad: if the ad speaks to someone fired after medical leave, the page answers that question first, then offers the consultation. A short form, a visible phone number, and no distracting navigation all help, as our guide to the lawyer landing page details.
The page carries compliance too. It must name the responsible lawyer or firm, keep the ad's promise without overstating it, and meet any state rule that applies to your website. OpenAI's crawler reviews it before approval, and it feeds ad selection, so a generic page hurts twice.
ChatGPT ads reach people who are actively working through a problem, with context no search keyword carries. Since August 2026, US firms licensed where their ads run can use them, after manual review by OpenAI. The format is fixed and labeled, the answer itself is not for sale, and the audience excludes paying subscribers. Costs are still unbenchmarked, which makes a measured test more useful than a large bet.
The firms that test now will know their real cost per signed case before the channel fills up. If you want a second opinion on whether ChatGPT ads fit your practice areas and your state's rules, talk to us about your firm.
Yes. OpenAI started testing ads in the US on February 9, 2026, for logged-in adults on the Free and Go plans. By October 21, 2026, ads ran in 89 countries. Plus, Pro, Business, Enterprise, and Edu accounts see no ads, and ads appear below the answer, labeled as sponsored, never inside it.

4.8 based on 30+ clients




Only 5
law firms per year
Position your law firm smartly in search results with our SEO and Google Ads expertise. See how we deliver stable, predictable results over time, all in an automated way.
