Paid ads · Google, Meta and more
Give ZeroTwo one brief and ask for several ad variants, each testing a different idea and each sized to the platform's character limits. This page shows the brief, a variant matrix with counted characters, and a test plan so you learn which idea works instead of guessing.
Free plan, no card required.
Variant sheet · illustrative
Google RSAStop chasing late invoices
26 of 30 characters, within the limitTests: Pain: late invoices cost time
Get paid closer to due date
27 of 30 characters, within the limitTests: Outcome: payment closer to the due date
Built for freelance designers
29 of 30 characters, within the limitTests: Audience call-out: name the buyer
Fictional product, copy written to show the format. Counts are computed from the text.
Over-length copy gets cut or rejected, usually in the middle of the hook. Put the limits in the brief, then count again in the platform's own editor.
| Platform and format | Headline | Body text | What to check |
|---|---|---|---|
| Google Search: responsive search ad | Up to 30 characters each. Up to 15 headlines, at least 3. | Up to 90 characters each. Up to 4 descriptions, at least 2. | Headlines and descriptions can appear in any order and Google may shorten text, so each one must make sense on its own. Display path fields allow 15 characters each. |
| Meta: Facebook Feed image ad | 27 characters recommended. | Primary text: 50 to 150 characters recommended. | These are Meta's recommendations rather than hard cut-offs. Check the placement preview, because Instagram, Stories and Reels placements have their own specs. |
| LinkedIn, TikTok, X, YouTube and others | Not listed here. | Not listed here. | Limits change by format and placement and were not re-verified for this page. Read the platform's own current ad specs before you put a number in the brief. |
Sources, opened 5 October 2026: Google Ads Help, About responsive search ads and Meta Ads Guide, Facebook Feed image ads. Platforms change specs, so confirm in the ad editor.
A generator can only be as specific as the brief. One benefit, one audience, one platform, and a request for variants that each test something different. Vague briefs return interchangeable lines.
Settling the message first helps; the brand messaging services guide covers that step, and the digital marketing assistant page shows how teams use ZeroTwo for wider paid-channel planning. This page is about paid ads only: organic captions belong on the AI caption generator.
Example brief · fictional product
This is the shape of output to ask for: a stated hypothesis, the copy, and the length against the limit. Variant F is one character over on purpose, to show the failure you would otherwise find in the ad editor.
| Variant | Idea it tests | Headline | Length (max 30) | Description | Length (max 90) |
|---|---|---|---|---|---|
| A | Pain: late invoices cost time | Stop chasing late invoices | 26 of 30 characters, within the limit | Late invoices mean hours of follow-up. Pocketledger sends the reminders for you. | 80 of 90 characters, within the limit |
| B | Outcome: payment closer to the due date | Get paid closer to due date | 27 of 30 characters, within the limit | Branded invoices and automatic reminders in one place, built for freelance designers. | 85 of 90 characters, within the limit |
| C | Audience call-out: name the buyer | Built for freelance designers | 29 of 30 characters, within the limit | Pick a template, add the project, send. See what is opened, due and paid at a glance. | 85 of 90 characters, within the limit |
| D | Effort: less admin per invoice | Invoice faster, chase less | 26 of 30 characters, within the limit | Turn a finished project into a branded invoice, then let reminders do the awkward part. | 87 of 90 characters, within the limit |
| E | Action: ask for the first invoice | Send your first invoice today | 29 of 30 characters, within the limit | Try Pocketledger on your next project and see every invoice's status in a single list. | 86 of 90 characters, within the limit |
| F | Mechanism: how reminders work | Late payments: automated nudges | 31 of 30 characters, 1 over the limit | Polite reminders go out before and after the due date, so you do not have to write them. | 88 of 90 characters, within the limit |
Illustrative copy for a fictional product, written to show the format. It is not output from a measured run and no performance is claimed for any line. Counts are computed from the text on this page.
Paste the brief, name the format, and ask for six variants with the idea each tests and a character count after every line. Then recount in the ad editor and trim anything over. Run the same brief on a second model if the first one keeps returning the same angle.
Once a concept has earned a bigger budget, you can expand an ad concept into a timed video script with the AI script generator, then brief the next round of headlines from what the test taught you.
A framework gives the model a spine to write against. None is proven better than another; pick by how much the reader already knows. Examples use the fictional Pocketledger.
Attention, Interest, Desire, Action
A default shape when the reader does not know you yet and needs a reason to keep reading.
Invoices piling up? Pocketledger turns a finished project into a branded invoice, then reminds your client for you. Send your first one today.
Problem, Agitate, Solution
When the pain is already felt and you can name it in the reader's own words.
Another invoice, another awkward nudge email. Chasing payment eats the hours you bill for. Pocketledger sends polite reminders so you can get back to design.
Feature, Advantage, Benefit
Considered purchases where the reader wants to see what the product does before the payoff.
Automatic reminders. You stop writing follow-ups. The payment conversation stays off your plate.
Useful, Urgent, Unique, Ultra-specific
Headlines and subject lines, where one line has to carry the whole message.
Late invoice? Send a reminder tonight.
Before, After, Bridge
Story-shaped ads where you can describe the reader's current mess honestly.
Before: a spreadsheet and three nervous emails. After: one list showing what is sent, opened and paid. Bridge: Pocketledger.
Variants only teach you something if the test is controlled. This is the plan to run alongside the matrix above.
What to record
Google's own figure, not ours
Google's help page says advertisers with one responsive search ad who add a second saw an average 6.6% increase in conversions at a similar cost per conversion. That is Google's internal data (15 to 20 August 2025) about running more than one ad per group. It says nothing about AI-written copy.
Angle, headline benefit, proof point or call to action. If two things change between variants, a difference in results cannot be traced to either.
Same audience, placement, budget split, schedule and landing page. A new landing page is a second variable.
Choose the metric that matches the campaign goal, and decide how long or how many conversions you will wait for before you look. Clicks alone can reward copy that attracts the wrong people.
Use the ad platform's own experiment or A/B test feature so it divides traffic for you, instead of swapping ads by hand and comparing different weeks.
Write down what you learned, keep the stronger variant, and make the next test change the next variable.
Which model writes your best headline depends on your product, voice and platform, and no published ranking has tested yours. Send the same brief to two or three models and score what comes back.
ZeroTwo gives you 60+ models from OpenAI, Anthropic, Google, xAI and others in one workspace. Browse them on the models page or start from chat with AI.
Count it in the platform's own editor. Language models often miscount characters.
Read it aloud next to your best existing ad. Paste three or four samples into the brief to anchor voice.
Every number, guarantee and superlative needs a source you could show the platform's reviewer.
If two variants make the same point in different words, you are not testing anything.
Restricted categories and claims vary by platform. Read the policy, not the model's assurance.
A generator drafts text. The advertiser answers for what runs. The wider brand visibility in AI search playbook covers how a campaign gets cited back.
Drafting a campaign's worth of variants is ordinary chat. Plan credits decide how much of it you can do.
Compare every plan on the pricing page.
Six things: the product in one sentence, the single benefit that matters most, who the audience is and what they already know, the platform and format with its limits, the tone, and exactly what you want back (how many variants, each with a stated idea). Name a framework such as AIDA or PAS if you want one. The example brief on this page is a working template.
Put the limits in the brief, ask for the character count next to each line, then count again in the ad platform's editor, because language models often miscount. Google's responsive search ads allow up to 30 characters per headline and 90 per description, and Google may still shorten text when it serves the ad. Meta publishes recommended lengths rather than hard limits.
Change one idea at a time, keep audience, placement and landing page the same, and decide the metric and stop rule before launch. Use the platform's own experiment or split-test feature so it divides traffic for you. The test plan section on this page lists what to record.
It can imitate a style you show it. Paste three or four of your strongest existing ads or pages, ask it to describe the voice (sentence length, vocabulary, words you avoid), then ask for new copy in that voice. Read the output against your originals; style imitation is not the same as knowing your product.
This page does not rank models, because there is no test behind a ranking. ZeroTwo lets you send the same brief to models from several providers; run two or three on your own brief and score the outputs with the checklist above. See the models page for what is available.
Claims, prices and deadlines, platform policy, trademark and competitor mentions, regulated topics, and message match with the landing page. The pre-launch checklist on this page lists them. A generator drafts text; the advertiser is responsible for what runs.
You can start on the Free plan with no card: 100 free credits to start, 20 bonus credits every day you log in, one file upload per message and a limited selection of models and tools. Paid plans raise those limits: Plus is $14.99 per month and Pro is $29.99 per month. See the pricing page for current allowances.
Expand an ad concept into a timed video script.
Organic captions and hashtags, a different job from paid ads.
Short organic posts and repurposing for social channels.
Long-form drafts with a source-backed editorial workflow.
How marketing teams use ZeroTwo across the funnel.
The marketing toolkit in one place.
Writing the follow-up emails for the same campaign? The onboarding email sequence playbook applies the same short-form discipline.
ZeroTwo publishes this page and is one of the tools it describes. The Free plan needs no card.