Google Ads started showing advertisers a preview of AI Max's generated ad copy in early October 2026, before anyone has to switch text customization on. Enter a final URL in the Asset optimization section of a Search campaign and Google's AI writes up to 10 sample headlines and descriptions from that page, in about 30 seconds, with no commitment. That sounds like a straightforward win: see it before you buy it. The part worth getting right is what to actually check in those ten samples, because approving them on sight is not meaningfully different from not previewing at all, and the copy a performance marketing account runs is one of the few places AI-generated text creates real legal and brand exposure.
Google Ads now generates up to 10 sample AI Max headlines and descriptions from your landing page URL before you opt in to text customization, reported by Search Engine Roundtable on 9 October 2026. The preview is a sample, not a locked set of assets, so approving it doesn't fix what a live campaign will say later. Run the five checks below before you opt in, and for regulated accounts, confirm your text guidelines and disclaimers actually hold before trusting the toggle at all.
What the preview actually shows, and what it doesn't
The feature itself is simple. You're in the Asset optimization panel of a new or existing Search campaign, you add or confirm a final URL, and Google's AI reads that page and produces sample headlines and descriptions it would generate for text customization. Your text guidelines, the words and themes you've already told Google to avoid, filter the samples; if your guidelines are strict enough to block everything, Google prompts you to loosen them rather than showing nothing. You don't have to opt in to look, and looking creates no obligation.
What it doesn't do is more consequential than what it does. The ten examples are a sample of what the model can produce for that URL right now, not a fixed set it will keep using. A live campaign generates and tests new combinations continuously as it runs, pairs them with different landing pages if Final URL expansion is also on, and adapts to the query it's responding to. Approving ten headlines in a preview tells you the model is capable of writing something you'd accept. It does not tell you what every future ad will say.
| What the preview confirms | What it can't confirm |
|---|---|
| The model can produce on-brand copy for this specific URL | Whether it will keep producing on-brand copy for every query it matches later |
| Text guidelines successfully filter out named words or themes | Whether a claim technically outside your excluded list still misrepresents the offer |
| Sample headlines read naturally and aren't keyword-stuffed | Whether a live-served variant pairs cleanly with the landing page Final URL expansion actually picks |
| The feature works in your account's language (English, Spanish, Portuguese, Italian, French, German, Japanese and Dutch, among others) | Whether required legal or disclosure text survives into every generated variant without a disclaimer control configured |
That gap is exactly why a one-time glance at ten samples is the wrong way to use this. It's a decision-support tool for the toggle, not a guarantee about the feature.
The five-point audit to run on the preview
Before switching text customization on, read the ten generated samples against these five checks specifically. Most of the coverage of this feature treats it as a convenience; it's more useful read as a pre-launch compliance and brand gate.
- Unsupported claims. Does any headline state a price, guarantee, award or superlative ("fastest," "#1," "guaranteed") that your actual landing page doesn't substantiate? The AI is writing from what's on the page, so a vague or outdated page produces vague or inflated copy.
- Missing required wording. If your industry or platform requires specific disclosure language, is it present in every sample, or only some? Text guidelines can force words out; they're not built to force words in reliably across every generated variant.
- Brand voice drift. Read the samples next to your existing written ad copy. If the tone, formality or vocabulary is noticeably different, that drift will show up at scale once the feature is generating variants continuously rather than the ten you reviewed once.
- Redundant or stuffed phrasing. Headlines that repeat the same keyword across multiple assets in a single ad group read as spam to a human and can waste ad strength. Check whether the samples differentiate or just restate the same two or three phrases.
- Landing-page dependency. If Final URL expansion is also on, ask whether the claims in these samples still hold if the click lands somewhere other than the URL you previewed against. A headline promising a specific offer only works if the page it sends people to still makes that offer.
We treat the preview as a gate, not a sign-off: run the five checks above every time a client's account surfaces it, and keep a dated screenshot of the samples reviewed alongside the decision, since "we looked at it in October" is a thin compliance trail by Q1.
When to leave text customization off regardless of what the preview shows
A clean preview is a reason to proceed carefully, not a reason to skip the decision. The accounts where the calculus is genuinely different are the ones carrying legal or regulatory exposure in their ad copy, and the control that changes the answer is a feature most advertisers haven't configured yet: Google's text disclaimers tool, announced earlier in 2026, which keeps required wording attached to an ad even when Final URL expansion sends the click to a different page. Before that existed, advertisers who needed guaranteed disclosure language had to choose between Final URL expansion and text customization. Now the two can coexist, but only if the disclaimer is actually set up.
| Business type | Keep text customization on when... | Leave it off when... |
|---|---|---|
| Healthcare, finance, legal, insurance | Text guidelines and text disclaimers are both configured and tested against real sample output, not just enabled | Either control is unconfigured, or no one has reviewed a preview against your specific compliance language |
| Local services, home services, trades | Messaging latitude is wide and the preview samples read as accurate, specific and on-brand | Pricing or guarantee claims vary by job and the AI-written copy can't reflect that variability |
| E-commerce, retail | Product and pricing data on the landing page is accurate and current, so the AI has something true to draw from | Promotions and pricing change faster than your landing pages update, since the AI will keep writing from whatever is live |
| B2B, lead generation | Copy is largely functional and benefit-led rather than legally sensitive | Claims about outcomes, savings or results need a human sign-off before they go live, and that review can't scale to continuous generation |
The pattern across all four rows is the same: the preview earns you the right to make an informed decision, but the decision itself still rests on controls that exist separately from the preview, specifically text guidelines and, when Final URL expansion is active, text disclaimers.
Frequently asked questions
What does Google Ads' new text customization preview actually show?
After you enter a final URL in a Search campaign's Asset optimization section, Google Ads generates up to 10 sample headlines and descriptions written by its AI, based on that landing page, in about 30 seconds. You can view these before switching on text customization or launching the campaign, and your text guidelines (words or themes you've excluded) filter which samples are shown.
Do I have to opt in to see the preview?
No. The preview is designed specifically so you can evaluate sample output before opting in to text customization or Final URL expansion. You can view it, reject it, and leave the setting off, with no commitment created by looking.
Does approving the preview guarantee my live ads will say the same thing?
No. The preview is a sample of what the AI can produce for that URL right now, not a locked set of assets. A live campaign generates and tests its own combinations continuously, so approving ten examples today does not fix the wording your ads will actually show next month.
Should regulated businesses turn text customization on?
Only if you can enforce your required wording through text guidelines and, where Final URL expansion is also on, through Google's text disclaimers feature, which keeps mandated text attached even when the AI chooses a different landing page. Without both controls configured and tested, the safer default for healthcare, finance, legal and insurance accounts is to leave text customization off and write ad copy manually.
How is this different from the AI Max migration I may have already gotten in September?
The September 2026 migration moved eligible Search campaigns into the AI Max campaign type automatically; that part was not optional. Text customization is a setting inside AI Max, not the migration itself, and it stays a campaign-level toggle you can switch off after the migration completes. The new preview exists to help you decide that toggle, not to reverse the campaign-type change.
The takeaway
A 30-second preview is genuinely useful, but only as the first half of a decision. Run the five-point audit against the actual samples your account generates, not against the idea of the feature. And if your ad copy carries legal or regulatory weight, don't let a clean-looking preview substitute for confirming that text guidelines and the rest of your AI Max settings are configured the way the preview implied they were. You're still the one liable if the copy is wrong, not the preview tool.
Sources & further reading
- Search Engine Roundtable, "Google Ads Enables Text Customization Asset Previews"
- Google Ads Help, "Turn text customization on or off in Search campaigns"
- MediaPost, "Google Gives Advertisers 30-Second Automated Ad Previews"
- Search Engine Journal, "New AI Brief And Text Disclaimers Come To Google AI Max"