Telegram Ads creative patterns — what advertisers actually run

After indexing 50K+ Telegram Ads creatives over four years, certain patterns recur far more than others. What this article can tell you is what advertisers repeatedly choose to run; what it cannot tell you is which of those choices performed, because Telegram exposes neither clicks nor conversions to third parties. Recurrence is a real signal — it reflects where experienced buyers put their money — but it is not proof of effectiveness.

#Copy structure that recurs

The 160-character limit forces brevity. The most frequent structure is:

1. **Hook line** — concrete claim or curiosity gap, 40–60 chars

2. **Value-prop** — what the reader gets and why now, 40–60 chars

3. **CTA tease** — short instruction matching the button, 30–40 chars

A typical example of that pattern in finance:

«Track 50 DeFi positions in one dashboard. Free for first 100 wallets. Tap to claim →»

Avoid:

• Lead with brand name — readers don't care about your brand until they care about the value
• ALL CAPS — Telegram moderation flags + reader fatigue
• More than 2 emojis — usually triggers «excessive emoji» rejection
• Questions you don't answer — «Want to earn more?» wastes the hook line

#Banner image patterns

Banner patterns that recur most often:

• **Product screenshot** with one bold value-prop overlay — readers see «what they get»
• **Clean brand mark** on solid colour background — high recognition, less fatigue
• **Result/proof imagery** — chart showing improvement, customer count, big number
• **Faces (real people, not stock)** — humans engage with humans, but only if photo looks authentic

Banner patterns we see far less often, and which experienced buyers avoid:

• Stock photos that look like stock photos (smiling models, generic offices)
• Heavy text-on-image (Telegram already shows your text below)
• Multiple competing visual elements (3+ products in one banner)
• Logo-only with no context

Banner spec: 1280×720, JPEG or PNG, under 1 MB. Light backgrounds are the more common choice in the creatives we index. Which theme a given reader sees is a client-side setting we cannot observe, so treat this as a convention among advertisers rather than a measured result.

#CTA button copy

The button is a 30-character slot. Which wording converts better we cannot measure — Telegram exposes neither clicks nor conversions to third parties. The order below is a working hypothesis from practice, not a measurement from our data:

1. **Action + reward** — «Get my $50 credit», «Claim free month»

2. **Action verb only** — «Subscribe», «Try it», «Download»

3. **Brand verb** — «Open Hamster», «Visit Catizen»

4. **Generic CTA** — «Learn more», «Find out»

Avoid: «Click here», «Submit», passive verbs.

#Targeting + creative match

What we can actually see is where each creative ends up — the thematic closeness of creative and channel:

• Finance creatives cluster heavily in finance channels rather than general-news ones — advertisers concentrate their buying where the topic matches, which is the behaviour we can see
• In non-English markets, advertisers running localised copy alongside English almost always keep the localised variant in rotation longer — a signal about their own choices, not a conversion measurement
• Tight topic targeting on a few concrete topics rather than broad spread — a working hypothesis from buying practice, not a measurement from our data

If you do nothing else, tighten your topic targeting and write creative copy that mentions the niche by name.

#Creative rotation cadence

We cannot observe creative performance — clicks and CTR are not exposed to third parties, so no such field exists in our data. What we can observe is **how long a creative stays in rotation**: the first date we see it and the last. Most creatives leave rotation within two to four weeks of first appearance, which is consistent with advertisers rotating on that cadence — but that measures the advertiser's behaviour, not the creative's decay.

The standard cadence: launch 3–5 variants on day one, drop the two weakest by day 7 and refresh the rest every 14 days with copy or banner edits — which ones are weakest only your own ad account can tell you. Telegram publishes no quality score and no factor feeding one — see /wiki/telegram-ads-auction-mechanics for what the platform does and does not document.

#What to A/B test

Do not test everything. Start with the variables you can actually change and read in your own account:

1. **Hook line** — different angles for the same offer

2. **Banner imagery** — product vs. proof vs. face

3. **CTA copy** — generic vs. specific reward

4. **Targeting niches** — 3 different topic combos

Don't waste auction budget on:

• Punctuation variations
• Single-emoji variants
• Reorderings of the same words

#Cabinet-specific creative rules

Moderation is the same in every cabinet — the ad policies do not depend on how you pay. What differs is the audience you reach:

• **Euro cabinet** — lead with utility, not claims. Avoid currency symbols in copy except for the CTA button.
• **Gram cabinet (formerly the TON cabinet)** — you hand-pick the crypto channels, so the audience expects on-chain proof; mention tx volume, TVL, audit. Avoid «moonshot» language.
• **Stars cabinet** — it only promotes your own channel or bot; if that bot runs a mini-app, its users are habituated to gamification, so game-language CTAs («Play now», «Tap to start») fit.

#See also

What other advertisers in your niche are doing right now: tgadsspy.com — search by niche, click into any advertiser to see their full creative library. Moderation rules: /wiki/telegram-ads-moderation. Launch playbook: /wiki/how-to-launch-telegram-ads.