A/B Testing Telegram Ads Creatives: How to Do It

An A/B test in Telegram Ads means running several versions of an ad on the same audience with the same bid, so you can use data to pick the version with the best cost per click or per target action. The ad account has no built-in split test, so each variant is launched as a separate ad. Below is how to frame hypotheses, how many impressions to wait for and which metrics to decide on.

Does Telegram Ads have built-in A/B tests?

No. Telegram Ads has no tool that splits the audience between variants and picks a winner for you. Back in 2022, agencies surveyed by eLama named convenient A/B testing among the features they missed in the platform. AdsGram’s guide (August 2026) recommends running variants as separate parallel ads.

Two features of the account directly affect testing:

• Targeting can’t be edited: according to the ads.telegram.org help page, it can’t be changed after an ad is created. For a new variant, use “Create a similar ad”, which copies the text and settings.
• Edits trigger moderation: according to AdsGram, changing the text or media sends an ad back for review, while changing the bid or daily budget does not. Editing one ad several times mixes up its statistics.

What to test in an ad

A Telegram ad is short — up to 160 characters of text, a button and optionally one media file — so there are few elements to test. The key rule: change one variable per test, or you won’t know what worked.

• Offer and angle: a benefit, a pain point, a question or a statement. This usually makes the biggest difference.
• Opening phrase: in the feed, users see the beginning of the text first.
• Media: text only versus a version with an image or video. Keep the CPM markups in mind; see our article on video ads in Telegram Ads.
• Emoji: regular or custom. A custom emoji raises the bid by 20%, so test it separately.
• Landing page: a channel versus a bot or website with the same text.

AdsGram advises testing different claims rather than synonyms of the same phrase: swapping “fast” for “quick” rarely produces a measurable difference. Ideas for angles are in our article on Telegram Ads creatives.

How to run an A/B test: step by step

1. State a hypothesis. For example: “Copy with a specific number will get a lower CPC than generic copy.”
2. Fix the audience. The same set of channels or topics, and the same geo and language, for all variants.
3. Set the same bid and budget. Otherwise the auction will show variants to different audiences in different volumes.
4. Create variants with “Create a similar ad”, changing only the element under test.
5. Name ads consistently to avoid confusion, for example project_hypothesis_targeting_date.
6. Leave the campaign alone until the variants collect enough impressions.
7. Review the results by cost per click and per target action, switch off weak variants and test the strong one on new audiences.

Moderation doesn’t approve every variant, so launch extras. The agency Digit told eLama (2022) that it uploads 7–10 variants per channel so that at least 3–4 are approved. How to lower the rejection rate is covered in our article on Telegram Ads moderation.

How many variants and impressions you need

Telegram doesn’t give an exact formula, but specialists agree on the order of magnitude. Artem Sudarikov of Magnetto.pro (ppc.world, August 2025) proposes a two-stage approach:

According to him, you shouldn’t draw conclusions from the first 100–200 impressions: each variant needs at least several thousand impressions, which usually takes 2–3 days. In 2022, Digit allowed 1–2 days, sometimes up to 3–4 days, for each targeting-and-creative pair, because results shift over the course of a day.

Keep the basic arithmetic in mind: with a CTR of about 1%, a variant with 5,000 impressions gets around 50 clicks. With a sample that small, a difference between 1.0% and 1.2% may well be random, while one between 0.5% and 1.5% is unlikely to be.

Which metrics decide the winner

CTR is handy as an early signal, but the decision should be made on money. A high CTR with expensive clicks or poor conversion often means untargeted traffic. Specialists advise looking at the whole funnel: CPM → CTR → CPC → CPA.

• CPC — what a click costs; shows how the copy performs in the auction.
• CPS or CPA — the cost of a subscriber, bot start or lead; this is the main test metric.
• Traffic quality — unsubscribes and the activity of new users in the first days.

For websites, conversions are tracked with the Telegram Ads pixel or UTM tags; for bots, with start parameters. More on the numbers in our article on Telegram Ads metrics.

Common testing mistakes

• Different audiences per variant: then you’re comparing channels, not creatives.
• Different bids: the ad with a higher CPM gets more impressions and a different audience.
• Edits during the test: every text change means another review and mixed statistics.
• Drawing conclusions too early: a few hundred impressions don’t give a reliable result.
• Testing with the whole budget: Magnetto.pro advises setting aside 30–50% of the planned budget for testing hypotheses and using the rest to scale the winners.

A winner also wears out over time: when CTR starts to fall while frequency rises, it’s time for a new round of tests. How to expand successful combinations is covered in our article on scaling Telegram Ads campaigns.

Tools that make testing easier

With dozens of variants, tracking each one by hand is impractical. A “targeting — creative — result” spreadsheet, a consistent naming scheme and Telegram Ads automation services that collect ad statistics and let you set rules all help; one example is the free TG Adviser. A selection of tools for working with the ad account is in our Telegram Ads services section.

FAQ