Skip to main content

ChatGPT will start rewriting your ads on October 30. Here is how a California B2B team keeps every claim its own.

From October 30, 2026 ChatGPT ads rewrite your copy and landing URLs by default. What California B2B teams risk under state ad law, and how to stay in control.

Orange upholstered chairs around a light wooden conference table in an empty meeting room

You wrote one careful headline for your ChatGPT campaign. Legal read it. Product signed off on the feature claim. Next week a buyer could see a different headline, pointing to a different page on your site, in a language your team never reviewed. Nobody on your team will have approved that version, and no screen in your account will show it to you first.

This is not a mistake by your media buyer, and it is not your fault. OpenAI told advertisers that from Friday, October 30, 2026, it will switch on "creative enhancements" for every new and existing self-serve ChatGPT campaign unless you opt out. If you sell B2B software or services from California, the timing matters more than usual, because the state's false advertising law reaches statements you make or cause to be made from California to buyers in any state, and it judges you by what reasonable care should have caught.

The fix is not to stop testing ChatGPT ads. It is to decide on purpose which campaigns may be rewritten, to clean the landing pages the system reads from, and to measure pipeline from your CRM instead of trusting the platform's view. This article explains exactly what changes on October 30, how it can hurt a B2B pipeline in California, New York and Texas, and the setup that keeps your claims and your numbers your own.

Oct 30
2026: ChatGPT ad rewrites switch on by default for self-serve campaigns
$2,500
California civil penalty cap per violation of the false advertising law
83%
of AI-using B2B buyers say AI influenced their vendor choice (Semrush)
$3 to $5
OpenAI's suggested starting maximum bid per click for ChatGPT ads

What OpenAI changed in ChatGPT ads

OpenAI is turning on automatic ad rewriting, which it calls creative enhancements, as the default for self-serve ChatGPT campaigns at the end of October 2026. The system can write new headlines and descriptions, change the landing page a click goes to, translate your text and resize your images, and those versions can run without anyone on your side seeing them.

The default flips on October 30

Advertisers learned about the change by email. Search Engine Roundtable published the email on October 20, 2026. It says that starting Friday, October 30, OpenAI will enable creative enhancements for new and existing self-serve campaigns, and that the ads "can run automatically without requiring you to review or approve each version." You can opt out before that date or change the setting at any time in Ads Manager. Existing account and campaign opt-outs are kept, and a campaign setting beats the account default.

The date is not perfectly settled. OpenAI's help page on creative enhancements, updated around October 24, now gives the start as October 31, 2026. Treat October 30 as your deadline. If you want a campaign left alone, turn the setting off before then, not after.

What the system can rewrite

The help page lists four changes OpenAI may make. It may create new or updated headlines and descriptions "to connect your offering to the conversation or user interests." It may translate your ad text into another language. It may resize images to fit the format. And it may send the click to a different page on your website that it judges more relevant.

The source material is your existing ad, the content of your landing page and the conversation the ad appears next to. Variations can be prepared in advance or written in real time. The same page answers the question every compliance lead will ask: you do not review each enhanced version, and the product does not offer a preview of every variation. That is a sharp contrast with how the ads got approved in the first place. OpenAI's Ads FAQ says ads must be reviewed before they can serve, so the version that passed review and the version a buyer reads can now differ.

Why California B2B teams feel it first

ChatGPT ads are new, and B2B teams are the kind of advertiser that tests a new channel early. According to PPC Land's report on the launch, OpenAI opened its self-serve Ads Manager beta to all US advertisers on May 5, 2026, after a pilot that began on February 9 with impression pricing at $60 per thousand. Click bidding came with the self-serve launch, which made small tests affordable for a SaaS team in San Francisco or a services firm in San Diego.

California is also where the legal consequence is broadest. The state's false advertising statute follows statements made from California, not only statements seen there. A Los Angeles vendor whose rewritten ad runs in front of a buyer in Ohio is still a California advertiser for that purpose. The state section of this article sets out the statute and the penalties.

How automatic rewrites hurt a B2B pipeline

Automatic rewrites hurt a B2B pipeline in six specific ways: unapproved claims, wrong landing pages, unreviewed translations, weaker lead quality, measurement you cannot untangle and spend that moves faster than you expect. None of them is guaranteed, and all of them are invisible until someone goes looking.

A claim you never approved runs under your name

B2B ads live on precise claims. "SOC 2 Type II." "Integrates with Salesforce." "Free for teams under ten." A rewrite built to match a conversation can soften, stretch or merge those claims, and OpenAI's own example shows the system adding context the original ad did not say. For a California company, the statute asks whether you knew or by reasonable care should have known a statement was misleading. Leaving an unreviewed generator switched on is a hard position to defend as reasonable care.

The click lands on a page you did not choose

Destination URL optimization means OpenAI may send the click to "a relevant page on your website." For a SaaS site that can be an old pricing page still indexed under a forgotten path, a careers page, a partner page, or a product page for a plan you retired. Your carefully built demo landing page, with its form, its proof and its tracking, gets skipped. The buyer arrives somewhere that was never designed to convert, and your cost per opportunity rises for reasons no report will name.

Translated copy that nobody on your team can read

Translation sounds harmless until a buyer clicks a Spanish or Korean ad and lands on an English page with English contract terms. A translated feature claim is still a claim. If your sales team cannot read the version the buyer saw, it cannot honor or correct it on the first call. A California seller with a national pipeline is exposed here, because the statute follows the ad out of the state.

The quick test: open every landing page your ChatGPT campaigns use and every page linked from them. Anything on those pages, including an old promotion, a retired plan or a draft case study, is now raw material for an ad. If you would not put it in a headline, take it off the page or opt that campaign out.

Lead quality can slip before anyone notices

The first operator reaction was not about compliance. PPC practitioner Melissa Mackey replied to the news on X that her team had already paused a ChatGPT ads test "due to poor lead quality." Rewrites tuned to relevance can widen who clicks. That helps a consumer brand chasing reach. It hurts a B2B team that pays for every click and only profits from the handful that become sales-accepted opportunities.

You cannot tell what the rewrites did

OpenAI is candid about this. The help page says performance changes over time "do not, on their own, establish the effect of creative enhancements." The FAQ adds that ChatGPT ads do not yet have performance benchmarks across advertisers or industries. So if your cost per lead moves in November, you will not know whether the rewrites, the season or the auction moved it, unless you set up a clean comparison before October 30.

Spend can run ahead of a cautious budget

The same FAQ explains that a daily budget is an average. With an unchanged budget, a campaign can spend up to twice its daily amount in one day and up to seven times it in a Sunday to Saturday week. OpenAI's basics page also warns that spend insights may lag by 7 to 8 hours. Put those together with new rewritten variations and a test budget can move quickly in the first days of November.

How the risk differs in California, New York and Texas

The rewrite feature is identical in every state, but the legal cost of a bad rewrite is not. California's false advertising law follows the advertiser, New York's newly expanded consumer protection law gives its Attorney General wider tools, and Texas lets many small business buyers sue a seller directly.

California: the law follows the advertiser

California's main false advertising statute is Business and Professions Code section 17500. It makes it unlawful to make or disseminate, or cause to be made or disseminated, an untrue or misleading statement about services, including over the Internet, either before the public in California or from California before the public in any state. The standard is not only what you knew. It is what "by the exercise of reasonable care should be known." A criminal violation carries up to six months in county jail or a fine of up to $2,500.

The civil side is where the exposure multiplies. Under section 17536, each violation can draw a civil penalty of up to $2,500, and the case can be brought by the Attorney General, any district attorney, a county counsel or a city attorney. That puts the San Francisco City Attorney, the Los Angeles City Attorney and the San Diego District Attorney all in the frame, not just Sacramento. The statute does not say how a court would count violations for an ad that rewrites itself per conversation, and that open question is not one you want to answer as the test case.

For a Bay Area SaaS company, a Los Angeles agency or a San Diego biotech supplier selling to labs, the practical rule is simple. Every claim a ChatGPT ad could make should already be a claim your team has approved, on a page you control. The California market page covers how that plays out across San Francisco, Los Angeles, San Diego and San Jose.

StateMain law on misleading adsWho can enforceMoney at stakeWhat to fix first
CaliforniaBus. and Prof. Code 17500, covers ads made from California to any stateAttorney General, district attorneys, county counsel, city attorneysCivil penalty up to $2,500 per violation (17536)Opt out until every landing page claim is approved
New YorkGBL 349, expanded by the FAIR Business Practices Act from February 17, 2026Attorney General for unfair or abusive acts; private suits for deceptive actsDepends on the case; AG named limited English proficiency and hidden pricing as targetsTurn off translation risk and show full pricing on linked pages
TexasDeceptive Trade Practices ActBusiness buyers with under $25 million in assets can sueUp to three times economic damages if the seller acted knowingly (17.50)Match ad claims to contract terms for smaller deals

New York and Texas

New York rewrote its main consumer protection law for the first time in 45 years. The New York Attorney General's announcement says the FAIR Business Practices Act lets the state go after unfair, abusive and deceptive practices under GBL 349, and frames it as protection for small businesses as well as consumers. A WilmerHale client alert notes that the law took effect on February 17, 2026, that only the Attorney General can bring unfair or abusive claims, and that courts' existing "consumer-oriented" test survives, which may keep purely private contract disputes out of reach. The same alert lists the Attorney General's examples, including practices that take advantage of New Yorkers with limited English proficiency and practices that obscure pricing. An automatically translated ad, shown to a small business owner in Queens and linked to an English-only pricing page, sits uncomfortably close to both. For a fintech or adtech team in Manhattan or Brooklyn, translation is the setting to check first.

Texas reaches B2B sales more directly than most states. Under section 17.45 of the Deceptive Trade Practices Act, a "consumer" includes partnerships and corporations, except a business consumer with assets of $25 million or more. Section 17.49 removes claims on written contracts above $100,000 where the buyer had independent counsel, and on any deal above $500,000. Section 17.50 allows up to three times economic damages when the seller acted knowingly. That means the exposure sits in the mid-market deals Austin, Dallas and Houston SaaS teams sell every day: a small firm that bought because an ad promised a feature, signed an order form without a lawyer and found the feature missing.

Who actually sees a ChatGPT ad

A ChatGPT ad reaches logged-in adults on the free and lower-priced tiers, not the paid business seats many of your buyers use at work. That shapes who clicks, what the rewrites optimize for and how much the channel should get from your budget.

Your buyers on Business plans see no ads

OpenAI's basics page for advertisers says ads are not shown to users on Plus, Pro or any Business plan, or to accounts identified as belonging to people under 18. A procurement lead researching vendors on a company ChatGPT seat will not see your ad at all. The same person asking from a personal free account on a Sunday evening might. For B2B, that makes ChatGPT ads a channel for founders, small firm owners and individual champions rather than the full buying committee.

AI already shapes the shortlist

The organic side of AI answers matters more than the ad slot. A Semrush survey of 519 US professionals who use AI at work, reported by PPC Land in July 2026, found that 66% regularly use AI to research vendors, 92% say AI has shaped their shortlist and 83% say it influenced their final choice. Only 7% said brand recognition decides whether they notice a vendor in an AI answer. Precise fit to the stated use case wins. That is the same signal the rewrite system chases, which is why your pages, not your ad copy, now carry the weight.

The principle: your landing pages are now your ad copy. Write them so any sentence lifted out of them is true, current and specific, and the rewrites, the AI answers and your buyers all read the same story.

Clicks from AI are still small next to search

Keep the channel in proportion. PPC Land's October 23 roundup reports a Semrush study that found Google AI Overviews on 88% of 400 tracked US commercial keywords in banking, software and ecommerce, and a separate review of 100 sites by Marie Haynes in which ChatGPT sessions were 0.62% of Google organic volume, with Google driving 96.5% of conversions. ChatGPT influences B2B buyers heavily. It does not yet send them in large numbers.

Opt out, or keep rewrites on with guardrails

Opt out of creative enhancements by default for any B2B campaign that makes specific product, security, pricing or compliance claims, and keep them on only for a deliberate, measured test. The decision is per campaign, because OpenAI lets a campaign setting override the account default.

When opting out is the right call

Turn rewrites off where a wrong word costs more than a better click-through rate earns. That means campaigns for regulated buyers, campaigns that mention certifications or integrations, anything with pricing, and any campaign whose landing pages you have not audited. It also means accounts where your legal team signs off ad copy. A sign-off process that the platform skips is not a process.

When a guarded test makes sense

Leave rewrites on for one awareness campaign that points to a page with plain, approved claims and nothing else, and run it against an identical campaign with rewrites off. Use the same context hints, the same budget and the same bid. OpenAI suggests starting maximum bids of $3 to $5 per click, so a controlled pair of campaigns does not need a large budget to show you the difference in sales-accepted opportunities.

Write for the format before the machine does

Short, specific copy leaves less room for drift. PPC Land's launch report says OpenAI recommends a headline of roughly 16 characters and a description of roughly 32. Copy that already fits, says one concrete thing and points to one purpose-built page gives the system little reason to rewrite and little material to rewrite with.

What works instead

What works is a closed loop: controlled settings, landing pages built as approved claim libraries, and pipeline data from your CRM that tells you which version of the channel actually produces opportunities. Each piece removes a way an automated system can act on your behalf without your knowledge.

Set the account default, then the campaigns

Before October 30, set the account-level default to off and switch rewrites on only where you have decided to test. Because campaign settings override the account default, a new campaign then starts safe and stays safe unless someone chooses otherwise. Write the choice down in your launch checklist, next to bid and budget.

Treat landing pages as the source of truth

OpenAI's help page tells advertisers to keep ads and landing pages current, including prices, promotions, product details and availability. Take that literally. Remove expired offers, retire old plan pages or redirect them, and keep each campaign's destination limited to pages where every sentence is approved. If the rewrite system may pick a "relevant page" anywhere on your site, the whole crawlable site is in scope.

Tag every click and keep your own record

OpenAI says UTM parameters persist on ad clicks. Give each campaign and each landing page its own tags, so your analytics and CRM show where a lead really arrived, including pages you did not choose. Save screenshots of what you can see in the account each week. If a buyer quotes an ad back to your sales team, you want a record of what you approved.

Feed the CRM back into the decision

Platform conversions count form fills. Your pipeline counts opportunities. Join the two through your CRM, and judge the rewritten and unrewritten campaigns by cost per sales-accepted opportunity, not by click-through rate. That single change turns a platform default into a business decision.

How Leadnox helps

Leadnox helps B2B teams test new channels like ChatGPT ads without handing their claims or their reporting to the platform. The work starts with an audit of what each campaign could say after October 30 and where each click could land.

Campaign and settings audit

We review your ChatGPT, Google and LinkedIn accounts for automated creative settings, map every landing page and linked page each campaign can reach, and flag claims that would be risky in a rewrite for buyers in California, New York and Texas. You get a short list: which campaigns to opt out, which pages to fix and which test to run.

Landing pages built as approved claim libraries

We rebuild demo, pricing and feature pages so every sentence is current, specific and cleared by your team, which makes them safer for automated ads and stronger for AI answers at the same time. This is how our B2B marketing work keeps paid media and AI search pointed at the same message.

Pipeline reporting from your CRM

We connect campaign tags to CRM stages so you can see cost per sales-accepted opportunity for each campaign, rewritten or not. That gives you the evidence to scale a channel, or to cut it, without guessing.

How to measure whether your setup is working

Measure ChatGPT ads with four numbers that come from your own systems: cost per sales-accepted opportunity, landing page match rate, lead acceptance rate and claim review coverage. Platform metrics describe delivery. These describe whether the channel helps your pipeline without adding risk.

Cost per sales-accepted opportunity

Divide spend by the opportunities sales accepts, per campaign, over at least a full month. Compare the rewritten and unrewritten campaigns side by side. Good looks like the rewritten campaign matching or beating the control on this number, not on clicks.

Landing page match rate

From your tagged sessions, count how many ChatGPT clicks arrived on the page you assigned to the campaign. With rewrites off this should be every click. With rewrites on, any click that lands elsewhere is a page you need to check, fix or remove.

Lead acceptance rate

Track the share of ChatGPT leads that sales accepts. A falling rate after October 30 is the early warning that relevance tuning is bringing in clicks your team cannot use, the same problem operators flagged the day the change was announced.

Claim review coverage

Keep a simple list of every page a campaign can send traffic to and whether your team has approved its claims this quarter. Good looks like full coverage for every campaign with rewrites on. Anything less is a page an automated ad may quote without your sign-off.

Frequently Asked Questions

Open ChatGPT Ads Manager, go to the creative enhancements setting and switch it off at the account level, then check each campaign, because a campaign setting overrides the account default. OpenAI says existing opt-outs are kept. Do it before October 30, 2026, the date in OpenAI's email to advertisers, even though the help page now lists October 31.

It may be. California's false advertising law covers statements a business makes or causes to be made, from California to the public in any state, that it knew or should have known were misleading. Leaving automated rewrites running unreviewed is hard to call reasonable care. Civil penalties reach $2,500 per violation. Ask your counsel how this applies to your campaigns.

They are worth a small, controlled test, not a large budget. Buyers on ChatGPT Business, Plus and Pro plans see no ads, so you reach individuals on free tiers rather than whole buying committees. Judge the test by cost per sales-accepted opportunity from your CRM, and run one campaign with rewrites off as a control.

Yes, when creative enhancements are on. OpenAI says it may optimize the destination URL to send people to a relevant page on your website. That can bypass your demo landing page and land buyers on older or unrelated pages. Tag every campaign, review where clicks really land, and remove or redirect pages you would not want in an ad.

OpenAI does not require it, and with rewrites on you cannot. Its help page says you do not review or approve each enhanced version and there is no preview of every variation. The practical control is upstream: approve every claim on the pages the system reads from, or opt the campaign out until you have.

Do you know what your ChatGPT ads will say after October 30?

Leadnox audits and rebuilds ChatGPT, Google and LinkedIn campaigns for B2B teams in California, New York and Texas, with automated rewrites under control, claims your legal team has cleared and pipeline reporting from your CRM. Start with a free audit.

Get My Free Paid Media Audit