Create

AI Assistant for Social Media Posts: Captions, Hashtags and Rewrites

Updated October 27, 2026 · Multi Post Social guide

What the AI assistant is, and where it shows up

An AI assistant for social media is only useful when it sits where the work happens. In Multi Post Social you will find a small sparkle button in three places: inside the composer next to the text box, on each card of the Create board, and under every comment or message in the social media inbox. There is also a full-screen AI Assistant page for longer sessions, such as planning a month of posts from one briefing.

The assistant knows the context it is opened from. In the composer it sees the channels you selected, so it respects the character limit of each one. On the ideas board it sees the title and notes of the idea. In the inbox it sees the comment, the post the comment is about and your saved replies. That context is why the drafts feel specific instead of generic.

Everything the AI assistant writes lands in an editable box. Nothing is queued, scheduled or sent until you press a button that says so. Treat it like a fast first draft from a colleague who has read your brief but not your brand book: good raw material, always worth a read.

Generate captions with three different hooks

The most used feature is Generate captions. You type a short brief, for example “we moved our weekly planning to Monday mornings and saved 20 minutes,” pick a tone and press Generate. The AI assistant returns three captions. Each one opens with a different kind of hook: a question, a bold statement, or a short story. The body carries your point in two or three short sentences. A call to action closes it, matched to the network: “save this” for Instagram, “what would you add?” for LinkedIn, a plain link for X.

Under each option you get three small buttons: Use, Refine and Regenerate. Use copies the caption into the composer. Refine opens a one-line box where you can type a correction such as “mention the free template” or “less formal.” Regenerate asks for three new options with the same brief. Most people find a usable draft within two rounds.

Hooks you can ask for by name

• Question hook: opens with a question your audience asks itself.
• Contrarian hook: challenges a common habit, then explains.
• Number hook: leads with a specific figure from your brief.
• Story hook: one sentence of scene, then the point.
• Direct hook: states the benefit in the first line, no setup.

Tone presets and per-network length

Tone is a preset you pick once per workspace and can override per post. Presets include Friendly, Professional, Playful, Direct and Educational. You can also write a custom tone note of up to 300 characters, such as “we are a small bakery, warm and local, never salesy,” and the assistant applies it to every draft. Teams on the Team plan and above can lock the tone note so contributors cannot change it.

Length follows the channel. When your post targets several networks, the AI assistant writes the base version at the shortest limit that still carries the idea, then offers one-click adjustments per tab: Shorten for X, Expand for LinkedIn, Make it a thread for Threads or Bluesky. Each adjustment keeps the facts in your brief and changes only the wording and structure. The character counter in the social media scheduler updates live, so you can see when a version fits.

Hashtag suggestions that match the text

Hashtags are suggested from the caption itself, not from a fixed list. Press Suggest hashtags and the assistant proposes two groups: broad tags that describe the topic and specific tags that describe the angle. Each tag shows a rough size label (large, medium, niche) so you can mix reach with relevance. Tap a tag to add it; tap the group name to add all. For Instagram the assistant offers to place them in the first comment instead of the caption.

If you already keep hashtag groups in the composer sidebar, the AI assistant checks them first and reuses tags you have approved before. You can save a suggestion as a new group with one click. The free hashtag generator on this site uses the same approach without an account.

Repurpose one idea into many formats

Repurposing is where an AI assistant saves hours rather than minutes. Paste a long piece of content, such as a blog section, a transcript or a newsletter paragraph, and press Repurpose. You get a set of drafts in different formats: a carousel outline with one point per slide, a thread of five to eight parts, a short LinkedIn post, a one-line X post with a link, a Pinterest title and description, and a YouTube Shorts script of about 45 seconds. Each draft keeps your wording for facts and quotes and changes only structure and length.

The output arrives as a group of posts in the composer, one per channel tab, so you can edit each one and add them to the queue together. The repurposing guide on our blog walks through a full example with timings.

Reply suggestions in the inbox

Under each open item in the inbox there is a Suggest reply button. The AI assistant reads the comment, the post it belongs to and the sentiment label, then drafts a reply in your tone. For a thank-you it suggests a short warm answer. For a question it answers from the post text and your saved replies, and flags the parts it is not sure about with square brackets, so you never send a guess. For a complaint it drafts an apology and a hand-off to support without promising anything specific.

The free plan includes five suggestions per week; paid plans have no limit. Suggestions are never sent automatically. The inbox guide explains how suggestions work together with labels, assignments and saved replies.

Working with the ideas board

On the Create board every idea card has a sparkle button. Press it and the AI assistant turns the title and notes into a first caption for the channels you use most. The idea moves to In progress and the draft is attached to the card. You can also ask the assistant for ideas: give it a theme such as “behind the scenes of our packaging” and it proposes ten angles as new cards in the Unassigned column. Delete the weak ones, keep the rest, and your board is full for two weeks.

What the AI assistant never does

A tool that writes for you needs clear limits, and ours are simple. The assistant never publishes, schedules or replies on its own; every action that reaches a network needs your click. It never invents facts, numbers, quotes or product details that were not in your brief; when a draft needs a fact you did not provide, it leaves a bracketed placeholder. It never changes the meaning of a quote when it shortens text. It does not write in a named person’s voice or impersonate another brand. It does not generate content that breaks our acceptable use policy, and it adds a reminder to disclose paid partnerships when your brief mentions a sponsor.

Prompt tips that get better drafts

The quality of a draft tracks the quality of the brief. You do not need to write long prompts; you need to include the right details. The table below lists the five details that make the biggest difference, with an example for each.

Three briefs and what came back

Abstract advice only goes so far, so here are three real-shaped briefs and the kind of draft the AI assistant returns for each. The drafts are shortened; the point is the shape, not the exact words.

Brief one: a small product update

Brief: “Calendar month view now shows the channel color on each chip. For social media managers. Tell them to open the calendar. No emoji.” Tone: Professional. Network: LinkedIn. The first option opened with a direct hook, “You can now see every channel at a glance in month view,” followed by one sentence on why color helps spot gaps and a closing line inviting people to open the calendar. The second option used a question hook about scanning a month in five seconds. The third told a two-sentence story about a Monday planning session. All three stayed under 600 characters and none mentioned pricing, because the brief did not.

Brief two: a behind-the-scenes post

Brief: “We photograph every product on the same wooden table from 2019. Our followers like our process posts.” Tone: Friendly. Network: Instagram. The drafts led with the table, each with a different angle: its age, the number of products it has held (left as a bracketed placeholder because the brief gave no number), and a question about readers’ own rituals. The hashtag suggestions grouped three broad tags about small business and three specific tags about product photography, with an offer to move them to the first comment.

Brief three: a shortened repost

Brief: a 900-character LinkedIn post about how often to post, with the instruction “shorten for X, keep the number.” The AI assistant returned a 240-character version that kept the one figure from the original, dropped the examples and ended with the link. A second press of Make it a thread produced four parts that restored the examples as separate posts. That is the normal flow: shorten first, then expand into a thread only when the short version feels thin.

Step by step: your first AI-assisted week

1. Set your tone. In Settings, choose a preset and write a custom tone note of two or three sentences about your brand voice.
2. Fill the ideas board. Open Create, press the sparkle button, give a theme and keep the best six of ten proposed angles.
3. Draft captions. Open one idea, generate three captions, refine once, and press Use. Repeat for five ideas.
4. Adjust per network. In the composer, use Shorten for X and Make it a thread for Threads. Add suggested hashtags to the Instagram first comment.
5. Add to the queue. The posts take the next slots in your posting schedule. Check them on the calendar.
6. Answer with suggestions. Each day, open the inbox and use Suggest reply for questions, editing each one before you send.
7. Review what worked. On Friday, look at the top posts in analytics and paste the best caption into your next brief as a reference.

Plan limits and privacy of prompts

The free plan includes 20 generations a month and five inbox reply suggestions a week. One generation is one press of Generate, Refine, Regenerate, Repurpose or Suggest, no matter how many options come back. Essentials, Team, Agency and Enterprise plans have no limit on generations. The pricing page has the full table.

Your prompts and the drafts are stored in your workspace so you can reuse them, and you can delete them from Settings at any time. Prompts are sent to the model provider only to produce the draft and are not used to train models. The assistant never reads your connected network passwords; it only sees the text and context you open it from. The security guide lists what is stored and for how long. If your organization needs prompts to stay in a specific region, the Enterprise plan supports that.

Using the AI assistant on your phone

The sparkle button is in the mobile app composer and inbox too. On a small screen the three caption options appear as swipeable cards; swipe to compare, tap Use. Dictate a brief instead of typing it: the microphone button turns speech into the brief text. Reply suggestions in the inbox are a single tap away, which makes a ten-minute inbox session on the train realistic. If you are on mobile data, drafts are small text payloads and arrive in a second or two.