When someone comments on one of your posts, your Clone answers using the post itself as context: the title, the copy you wrote and — for videos — a transcript of what's spoken. What it can never do is look at the pixels. It doesn't see the photo, the video frames or any text burned into the artwork. So the single highest-leverage habit for public posting is simple: whatever the visuals say that your words don't, put it in the copy.
What your Clone sees when it replies
Every time a comment arrives on a Facebook, Instagram or YouTube post, your Clone is briefed with the post's context before it writes a reply. Here's exactly what makes it into that briefing — and what doesn't:
| Post element | Does your Clone see it? |
|---|---|
| The post title and copy (caption / description) | Yes — always included |
| Words spoken in a video | Yes — transcribed automatically when video context is enabled |
| First-level comments you post on your own post | Yes — when own-comment context is enabled |
| The photo itself, or the video's frames | No |
| Text overlays and captions burned into the video | No — only the audio is transcribed |
The video and own-comment sources are governed by two switches in the Connection's Account options panel — on the Connections page, select the account's row and click the Settings (gear) button. Look for Transcribe and use post videos as context and Use account's first level comments as context; both are off by default, and for public posting we recommend turning both on.
The post's context is prepared once per post and reused for every comment on it — it isn't rebuilt or re-sent for each reply. Write the copy as complete as it needs to be.
Say in the copy what the picture shows
Your followers look at the photo; your Clone reads the caption. If a fact exists only in the image, the Clone simply doesn't have it — and the questions people ask under a post are almost always about what they're looking at.
Picture a post: a photo of someone at dinner wearing a pair of shoes you sell. The first comment is "What model are those shoes? 😍". If the caption reads "Ready for tonight ✨", your Clone has nothing to work with. If it reads "The Adora heels in cherry red — our favorite for a casual dinner", the Clone answers the model, the color, even the occasion, in your voice, instantly.
| The post | Caption that leaves your Clone blind | Caption it can answer from |
|---|---|---|
| Photo of a model wearing your shoes | "Ready for tonight ✨" | "The Adora heels in cherry red — perfect for a casual dinner. Sizes 22–27 in stock." |
| Reel of a barista pouring latte art, music only | "Morning ritual ☕" | "Pouring our new Kenya single-origin — roasted in-house, in the shop and online from Friday." |
| Carousel of an event space being set up | "Almost ready! 🎉" | "Setting up Saturday's wine tasting at our Roma Norte location — doors at 7pm, tickets in bio." |
Notice the good captions are still marketing copy — short, natural, on-brand. Naming the product, the variant and the practical details does double duty: it sells to the humans scrolling past and it arms your Clone for the comments.
Before publishing, look at the visual and ask: what's the first thing someone will ask about this? The model? The price? The date? The location? Make sure the copy contains that answer.
Videos: what the transcript covers — and what it doesn't
With Transcribe and use post videos as context turned on, the spoken audio of your videos is transcribed automatically and folded into the post's context. That changes what you need to put in the copy:
- Talking videos carry themselves. If you say it out loud — product names, prices, dates, how-to steps — the transcript delivers it to your Clone. Mentioning key names verbally is enough.
- Music-only Reels, Shorts and b-roll give the Clone nothing. No speech means an empty transcript, so the caption has to do all the narrating: what's shown, what it's called, where to get it.
- On-screen text is invisible. Titles, price tags and captions burned into the video aren't read — only the audio is. Anything important that appears on screen should be repeated in the copy or said out loud.
A promo code shown as a text overlay in the final frame of a Reel is invisible to your Clone. Someone asks "what was the code?" and it can't answer. Put the code in the caption too.
Add context after publishing
Posts age: sizes sell out, dates move, stock comes back. You have two levers that update your Clone's briefing without recreating the post:
Edit the copy
Updating the caption updates what your Clone reads. If the red sold out and only navy is left, say so in the copy.
Comment on your own post
With Use account's first level comments as context turned on, first-level comments made from your own account are handed to the Clone as additional context from the post owner. A comment like "Update: cherry red is sold out — navy and beige still available" shapes every reply from then on, and your followers see it too.
Quick checklist before you hit Publish
- Name things precisely. "The Adora in cherry red", not "these beauties". Model names, variants and colors are exactly what people ask about.
- Answer the predictable questions in advance — price, sizes, availability, where to buy, dates and locations for events.
- Repeat anything that lives only in the artwork — promo codes, on-screen prices, text overlays.
- For silent videos, let the caption narrate — assume the Clone hears nothing, because it doesn't.
- Keep it current with your own comments — sold out, rescheduled, restocked.
Frequently asked
No. It works from text: the post's title and copy, the transcript of a video's spoken audio (when video context is enabled), and your own first-level comments (when that option is enabled). It never analyses the image or the video frames — which is why the copy has to carry what the visuals show.
No — transcription covers the spoken audio only. On-screen titles, price tags and burned-in captions are invisible. Say the important parts out loud or repeat them in the copy.
No. The post's context is prepared once per post and reused for every comment on it, so a complete caption doesn't add per-comment work. Write what the post needs.
The post copy is always used — no setting involved. Video transcripts and own-comment context each have a switch in the Connection's Account options panel (Connections page → select the row → the Settings gear): Transcribe and use post videos as context and Use account's first level comments as context. Both are off by default.
Last updated July 23, 2026 · AI Clones
