A/B copy variants
Give it one ad, button label or subject line and get back versions that truly differ: one leads with a benefit, another with urgency, another with social proof, another with a direct question. Your test then compares ideas, not adjectives. You pay per use, with no subscription.
Run it online
Run this on our servers with your account. Free tools run in your browser; this one bills your KIT balance per the price above.
A/B testing only works when the variants are actually different from each other, not the same sentence rearranged three times.
Why A/B variants have to differ in substance
A split test only teaches you something when the versions say different things. Ask for three versions of a sentence and you often get the same idea with new adjectives, so whichever wins, you learn nothing. Here each variant is built on its own angle: one leads with a benefit, another with urgency, another with social proof, another as a direct question. That structural difference turns the test into a real comparison of ideas rather than of word choice. Every version stays within the facts you supply.
From direct mail to the checkout button
Split testing is older than online marketing: direct-mail copywriters were testing headlines against each other decades before the web existed. What changed online is speed. A test that once waited weeks for the post can now show a clear result in days, so the slow part is no longer waiting for answers but writing enough distinct versions to test. Typical uses are four ad angles before a campaign launch, the wording of a checkout button, three subject lines for an email send, or a landing page headline tested alongside its button.
Why split testing rewards this kind of structure
Split testing as a formal practice predates digital marketing entirely — direct-mail copywriters were running controlled headline tests with postage-stamp economics decades before the web existed. What changed online is the speed of the feedback loop: a test that once took a mail cycle of weeks can now report significance in days, which means the bottleneck shifted from waiting for results to producing enough genuinely distinct variants to test in the first place.
What you can do with it
Paid ad campaign launch
A performance marketer generates four distinct ad copy angles before a campaign goes live instead of manually rewriting the same line four times.
Checkout button testing
An e-commerce team tests different call-to-action phrasing on the checkout button without guessing which wording actually reduces cart abandonment.
Email subject line split test
A lifecycle marketing tool requests three subject line variants per campaign and auto-assigns them to test segments before the send.
Landing page headline test
A growth team runs a headline test alongside a CTA test in the same release cycle, generating both variant sets from the same product description.
FAQ
Is the A/B variant generator free?
There's no free tier — free tiers get abused and slow everyone down. No subscription, no tokens, no invented credits, and a failed task is never charged.
What happens if a variant generation task fails?
Failed tasks retry automatically up to three times, and if all attempts fail you receive a clear error with no charge for that task.
Why is each variant built on a different angle instead of new wording?
Because two versions that differ only by an adjective tell you nothing when the test ends. Each variant starts from its own angle, such as a benefit, urgency, social proof or a direct question, so the winner shows which idea worked.
Can a variant add a discount or a deadline I never mentioned?
No. Every variant stays within the facts in your source copy, so no version gains a discount, a deadline or a feature you do not offer. A winner built on an invented promise would teach you nothing you could use.
What kind of copy can I turn into variants?
Any short piece you plan to test: an ad, a button label, an email subject line or a landing page headline. You decide how many variants you want, for example four angles before a campaign goes live.
Related tools
For developers — API access
Everything on this page is available programmatically. This section is for teams who want to wire it into their own systems; everyone else can just use the tool above.
API endpoint
Prefer to automate it? One authenticated POST creates the task; the result comes back by webhook or a signed link. The same capability also runs here on the web, by email and from Telegram — and soon from our app too.
Call it from your stack
curl -X POST https://api.kit.forhosting.com/text/ab-variants \
-H "Authorization: Bearer $KIT_KEY" \
-H "Content-Type: application/json" \
-d '{"text":"El informe trimestral cerró con 1,2 M€ de ingresos."}'const res = await fetch("https://api.kit.forhosting.com/text/ab-variants", {
method: "POST",
headers: {
"Authorization": `Bearer ${process.env.KIT_KEY}`,
"Content-Type": "application/json"
},
body: JSON.stringify({
"text": "El informe trimestral cerró con 1,2 M€ de ingresos."
})
});
const { task_id } = await res.json();import os, requests
res = requests.post(
"https://api.kit.forhosting.com/text/ab-variants",
headers={"Authorization": f"Bearer {os.environ['KIT_KEY']}"},
json={
"text": "El informe trimestral cerró con 1,2 M€ de ingresos."
},
)
task_id = res.json()["task_id"]<?php
$res = file_get_contents("https://api.kit.forhosting.com/text/ab-variants", false, stream_context_create([
"http" => [
"method" => "POST",
"header" => "Authorization: Bearer " . getenv("KIT_KEY") . "\r\nContent-Type: application/json",
"content" => '{"text":"El informe trimestral cerró con 1,2 M€ de ingresos."}',
],
]));
$task = json_decode($res, true);body := bytes.NewBufferString(`{"text":"El informe trimestral cerró con 1,2 M€ de ingresos."}`)
req, _ := http.NewRequest("POST", "https://api.kit.forhosting.com/text/ab-variants", body)
req.Header.Set("Authorization", "Bearer "+os.Getenv("KIT_KEY"))
req.Header.Set("Content-Type", "application/json")
res, _ := http.DefaultClient.Do(req)Example request
{
"text": "El informe trimestral cerró con 1,2 M€ de ingresos."
}Example response
{
"task_id": "tsk_a1b2c3d4e5f6a1b2c3d4e5f6",
"type": "text.ab_variants",
"status": "queued",
"_links": {
"result": "/tasks/tsk_…/result"
}
}The API is asynchronous: the call returns a task_id immediately and the result arrives by webhook. Polling is capped at 1 req/s per task.
Pricing
Published price — no tokens, no invented credits. A failed task is never charged.
Limits
max_tokens | 20000 |
Errors
| HTTP | Code | Meaning |
|---|---|---|
401 | unauthorized | Missing or invalid API key. |
402 | insufficient_balance | Your balance doesn't cover the task price. |
404 | unknown_type | That task type doesn't exist. |
429 | rate_limited | Too many requests. Use the webhook instead of polling. |
422 | task_failed | The task failed after 3 retries. You are never charged for it. |
Technical notes
This endpoint takes one piece of copy — an ad, a button label, a subject line — and produces genuinely distinct alternatives built on different angles, so your test has something real to measure.
The problem with most 'variant' generators
Ask a generic model for three versions of the same sentence and you'll often get three versions that differ only in adjective choice — which tells you nothing when the test concludes. text.ab_variants is built specifically to diversify the underlying angle: one variant might lead with a benefit, another with urgency, another with social proof framing, another as a direct question. That structural difference is what makes a split test statistically meaningful instead of cosmetic.
How the request and result flow works
Either way, nothing blocks your application thread waiting on a model response. You send a POST to /text/ab-variants with your base copy and the number of variants you want, and get back a task_id right away while generation happens asynchronously. The finished set arrives at your webhook endpoint, or sits behind a signed link for 24 hours if you'd rather pull it yourself.
Copy that stays honest across every variant
Every variant is grounded in the facts you provide about the product or offer — the endpoint won't manufacture a discount, a deadline or a feature that isn't in your source copy just to make one version punchier. That matters doubly in A/B testing: a variant that wins on an exaggerated claim will lose on conversion quality downstream, and you'll have optimized for the wrong metric.
Fitting it into your testing pipeline
Because results land via webhook, you can wire this directly into a testing platform's variant-upload step, a CI job that regenerates ad copy before a campaign refresh, or a landing-page builder that needs three headline-and-button pairs before it publishes. Combine it with the headline endpoint when the test spans both the page title and its call-to-action button.
Technical questions
How do I generate A/B test copy with an API?
POST your base copy and desired variant count to /text/ab-variants, get a task_id back immediately, and receive the finished variants by webhook or a signed 24-hour link once processing completes.
How many variants can I request at once?
You specify the count in your request; there's no fixed cap built into the pricing model since each variant is billed individually at $0.0135.
Are the variants actually different, or just reworded?
The endpoint is built to vary the underlying angle — benefit-led, urgency-led, social-proof-led, question-led — not just swap synonyms, which is what makes them useful for a real statistical test.
Can I use this for email, ads and buttons in the same call?
Yes, submit whichever copy type you're testing; the endpoint works on any short-form marketing copy, from subject lines to button labels to ad headlines.
How is pricing calculated for bulk variant generation?
Each call costs a flat $0.003 request fee plus $0.0135 per variant returned, so testing ten campaigns with four variants each is a predictable, itemized cost.
Do I need to poll for results?
No — set a webhook URL and results are pushed to you automatically, or use the signed link if you prefer to fetch results on your own schedule within 24 hours.