Meta Rewrites Your Ads – and From November 20 You Are Liable for It

Imagine you run an ad on Facebook or Instagram. Text, image, offer: all checked and approved by you. The next day, a version is running that Meta’s AI has reworded, cropped or extended. You have never seen that version. And from a certain cut-off date, the new terms say that your company alone is responsible for what that version claims.

That is exactly what two reports by the trade outlet PPC Land, written by Luís Rijo and published on October 22 and October 23, 2026, are about. The reports draw on Adweek and on a Meta statement to Reuters. We read both articles in full. This post summarizes what they prove, what they explicitly leave open and what a company with a Meta ad account can prepare in the coming weeks. It is not legal advice, and it does not tell you whether to agree. It helps you make the decision consciously instead of letting a missed date make it for you.

What the sources prove: a new “AI Features” section

The first PPC Land article, dated October 22, 2026, is titled “Meta’s AI rewrites the ads. From October 30, advertisers carry the risk”. It reports that Meta’s revised Commercial Terms, according to Adweek, contain a new section called “AI Features”. Under this section, advertisers are “solely responsible” for anything Meta’s AI creates or changes in their ads. According to Adweek, this also applies to features that Meta itself switched on.

The section also acknowledges that these features can make mistakes and do not always deliver accurate, complete or suitable results. In other words, Meta writes into its own terms that the AI can be wrong, and places responsibility for the consequences with the advertiser.

Please keep two things apart:

• What is new: According to Adweek, an earlier limitation has been dropped. Previously, advertisers were reportedly liable for AI output only if they had used Meta’s AI creative tools themselves. Now responsibility is meant to rest with them even when Meta activated the feature.
• What is not new: The liability cap (the higher of 100 US dollars or the fees of the last twelve months) and the duty to indemnify Meta against third-party claims already existed before, according to PPC Land.

The source’s own caveat matters here: PPC Land could not check the contract passages against the original. The account relies on Adweek’s reporting. Meta, for its part, says this is a consolidation of existing rules. Whether advertisers would also be liable for features they never activated was left open by Meta, according to the article. We report it that way and add nothing.

When a platform writes into its terms that the AI can be wrong and at the same time decides that the customer is liable for it, then approval in your own house is no longer a formality. It is the only place where you can still have influence.

The delay: November 20 instead of October 30

The second article, dated October 23, 2026, is titled “Meta delays its AI terms to November 20. The trust problem does not move”. Its core statement: Meta is only moving the date. The revised Commercial Terms and the Self-Serve Ad Terms are now meant to apply from November 20, 2026 instead of October 30, 2026. A Meta spokesperson told Reuters this on October 23. As the reason, Meta said it needs three additional weeks to explain to partners and other stakeholders “what’s changing and what’s not”.

The article also records what the delay does not mean:

• The text is unchanged, according to PPC Land. Meta did not change the contract text. On the UK preview page, October 30 was still shown about seven hours after the Reuters report.
• There were complaints. Publishers and advertisers had complained to competition authorities in the EU and the UK. The European Commission welcomed the delay. However, according to the article, no authority has suspended or concluded a proceeding.
• The debate continues. Theo Bamber of the News Media Association wrote on LinkedIn: “This delay is welcome, but it is a pause not a solution.”

For you as an advertiser this means: you have three more weeks, but no new information about what will ultimately apply. The new cut-off date is a working window, not an all-clear.

“Continued use counts as acceptance”: what that means in practice

One point from the second article is easy to overlook in practice. After the notice, continued use of Meta’s services counts as acceptance. Nothing is signed. Whoever simply keeps campaigns running on November 20 (or on the day the terms apply to their account) has agreed under this mechanism.

According to PPC Land, partial acceptance does not exist. You cannot accept the advertising terms and reject the AI liability. Whoever objects has to stop commercial use of Meta’s services. For many businesses that is not a realistic option if Meta is one of their most important channels for customer inquiries. So the question is how you live with the clause, not whether you can get around it.

Our limit applies here as well: whether such a clause is valid under German or European law and what it means for your specific contract cannot be answered in this post. That belongs with your legal counsel. What you can do yourself is mark the date in your calendar and not leave the decision to chance.

Where the risks lie in practice: hidden switches and silent changes

The first article quotes several professionals who describe everyday life in ad accounts. Ryan Adams of the agency Exverus says: “Meta auto-applies the enhancements, and many of the options are tucked away in hidden or collapsed menus.”

Kayla Loecker of AdCellerant describes the consequence for liability this way: “shifting 100% of the legal liability onto the advertiser, Meta creates huge operational risks.”

The article also gives an example. According to PPC Land, Bookshop.org found that an ad had been “improved” by Meta without the company’s knowledge. The company then deactivated all AI features for ads. We do not reproduce what exactly was changed, because we could not verify it beyond the primary reports. The pattern is what matters: a change was noticed only after it was live.

And there is a second deadline that is closer than November 20. According to PPC Land, product browsing will be activated in existing Facebook and Instagram ads from October 28, unless advertisers have objected beforehand. Seen from today, that date is only a few days away. If you run catalog or product ads, it is worth checking this setting this week. We report only the date and the mechanism here; the source does not prove what effect the feature has on your results.

Availability: features are already active, the liability clause is coming

There is nothing to try out on this topic, but quite a lot to check. The status according to the sources:

• AI features in ads: According to the reports, they are in many cases switched on by default. The principle is opt-out: whoever does nothing is using them.
• Liability clause: After the delay, it is meant to apply from November 20, 2026.
• Product browsing: Activation from October 28, unless an objection was made beforehand.
• Region: The sources name no exception for particular countries. They also do not expressly prove the link to Germany, Austria and Switzerland. Whether and how the terms apply to your account, you can see in your ad account and clarify with your legal counsel.

We therefore do not frame this topic as “You can use this now” but as “This is coming for companies”. A contract clause with a fixed cut-off date is not a product you switch on, but a framework condition you prepare for.

What this means for a mid-sized company

First: approval moves from Meta to you. If Meta states in its terms that the AI can be wrong and that you bear the consequences, then internal control is the only protection. That is true regardless of how well the features perform on average. A single wrong price, a distorted product depiction or an unsuitable text is enough to cause trouble with customers or competitors.

Second: you need an answer to “Who is responsible?”. In smaller businesses, one person often looks after the ad accounts on the side, sometimes an external agency. Clarify who in your company is responsible for the settings in Meta, who sees changes and who decides in case of doubt. If an agency runs the account, a look at the contract also belongs here: who is liable there for changes the platform makes itself?

Third: documentation is your trail. If someone later asks why an ad looked the way it did, only what you recorded will help: which AI features were active at what time, who decided that and what you checked. A dated screenshot of the settings takes five minutes and can spare a lot of dispute.

Fourth: the mid-market has no special role, but it has an advantage. Large advertisers have legal departments and direct contacts at Meta. Small businesses do not, but their accounts are manageable. You can go through in one afternoon what a corporation has to coordinate over weeks.

Which teams at Amazing Agents would be responsible: Daniel, Nadja and Magdalena

On amazing-agents.com, this topic belongs to the Research & Analytics department. Agent Daniel is the analytics specialist there. The website describes that he collects performance data across channels, builds reports and flags anomalies before they become a problem. He delivers a weekly report with a recommendation for the following week. For the Meta question, that is the fitting frame: if an ad set’s metrics suddenly deviate, someone should see it early, and that is the job of analytics.

Agent Nadja leads the research team. According to the website, she delivers research briefings and runs competitor scans. For platform terms, that means someone watches what changes, puts it in context and passes it on as a short briefing before a deadline passes. The legal assessment stays with your legal counsel; a briefing does not replace a lawyer.

For the check before publication, the website names the agent Magdalena in the Tech & Ops department: no content goes live without her brand check. We would carry this idea over to your advertising. What appears in your name goes through a review before it goes live. When a platform makes changes afterwards, a second look at what is actually being delivered is needed.

We are describing an approach here, not a result from a client project. In an audit, we would define with you which changes to ads a human must see and approve, which metrics Daniel watches and which platform terms Nadja keeps an eye on.

What we do not know

Honesty is part of a good assessment. These points are not proven by the two sources:

• The exact contract wording: PPC Land could not check the passages against the original. The liability clause relies on Adweek’s reporting, and the Reuters report is not linked in the article.
• Liability for features never activated: According to PPC Land, Meta left this question open.
• The link to Germany, Austria and Switzerland: The sources name no country exception, but do not expressly prove the DACH link.
• The legal consequences: We have not checked whether the clause is valid in your case, whether it triggers obligations toward third parties and how it relates to your agency agreement.
• The effect on your results: Whether the AI adjustments improve or worsen your ads is not proven by the sources. You would have to compare your own data for that.
• What happens after November 20: Whether Meta changes the text by then or authorities step in is open.

What you should prepare now: five steps

Because this is about a clause with a cut-off date and not a product to switch on, the steps consist of preparation. Plan one afternoon for it.

1. Note the dates: Put October 28 (product browsing) and November 20, 2026 (new terms) in your calendar, with a reminder one week ahead and the question: who does what by when?
2. Check the settings in the ad account: Open your Meta ad account and go through the ad settings, including the collapsed menus. Note which AI features are active. Decide consciously which ones to keep and which to switch off.
3. Document the state: Record the status with a date, for example as a screenshot or short note, and keep it in a fixed place. Repeat this after every change.
4. Name a responsible person: Define who approves ad content, who checks the settings and who stands in. If an agency looks after the account, clarify in writing who is responsible for platform-side changes.
5. Involve legal counsel: Put the new terms in front of your legal counsel before the cut-off date is reached. Ask specifically what acceptance by continued use means for your account and whether you should change anything beforehand.

Agent with limits and an approval step: the difference

The Meta case shows in miniature what matters with every agent you bring into your processes. An AI that adjusts texts and images can be useful. What decides is what happens before the result reaches your customers.

Without an approval step: The AI changes the ad, and it runs immediately. The mistake is noticed only when a customer gets in touch, a competitor complains or the invoice arrives. By then the ad has already been delivered, and under the terms described, responsibility lies with you, without you ever having seen the change.

With an approval step: The AI proposes a change, and a human sees it before it goes live. The settings are documented, responsibility has a name, and you can explain what happened and why. An agent with limits and an approval step works within a frame that you have set, and puts the decisions that matter in front of a human.

On amazing-agents.com this frame applies to all agents: approval steps for critical actions, and emails are created as drafts, not sent directly. An ad in your name should be treated by the same principle: it is a draft until someone approves it. The delay to November 20 gives you the time to set up this frame for your ad accounts. Use it.