Individual prompting habits vary wildly. One person's prompts produce strong drafts; another's produce generic text. A shared library spreads the best practice across the team and turns personal know-how into a team asset. The Academy's glossary explains brand voice.
What to include for each prompt
| Field | Example |
|---|---|
| Name | Instagram caption, product launch |
| Purpose | Draft three caption options for a new product |
| Prompt text | The full prompt with placeholders in brackets |
| Inputs needed | Product details, audience, offer, example posts |
| Tool | Which approved AI tool to use |
| Tips | Follow-up prompts that improve results |
| Owner | Who maintains it |
| Last tested | Date and notes |
Organising the library
Group prompts by task: planning, writing, video, design, community, analytics and reporting. Use clear names and tags so people find prompts quickly. A shared document, spreadsheet or knowledge base works; complexity is not needed.
Core prompts to start with
- Content ideas within pillars.
- Caption drafts in brand voice.
- Hook variations for short video.
- Repurposing a long piece into platform posts.
- Reply drafts for common comments.
- Comment summaries for monthly reports.
- Creative briefs for designers or creators.
- Report summaries in plain English.
The Academy's guide to AI prompts for social media offers templates to adapt.
Building voice into prompts
Include a standard brand voice block at the top of writing prompts: personality, formality, banned words and spelling. Keep it in one place so updates apply everywhere. The Academy's guide to keeping AI content on brand explains what to include.
Platform constraints
Add platform limits to relevant prompts so drafts fit first time. For example, X posts are limited to 280 characters free; up to 25,000 for Premium subscribers✓, and YouTube titles to 100 characters (about 70 visible in search)✓. Update these when platforms change. The Academy checks platform facts weekly on its changelog.
Testing prompts
Before adding a prompt, test it several times with different inputs. Note where it fails and add instructions to fix those failures. Record example outputs so others know what to expect.
Version control
Prompts improve over time. Keep a simple version history, noting what changed and why. If results suddenly worsen, perhaps after a tool update, you can compare versions.
Safety notes
Add reminders to prompts that involve sensitive tasks: do not paste personal data, check facts, review claims against advertising rules and keep a person accountable. The Academy's guide to an AI policy for marketing teams covers wider rules.
Encouraging use
Introduce the library in a short session, show before and after examples and ask the team to suggest improvements. Recognise people who contribute strong prompts. Libraries that grow from team input get used.
Keeping it current
Review the library quarterly. Remove prompts nobody uses, update platform limits, refresh examples and retest prompts after major AI tool updates.
Sharing with agencies and freelancers
If agencies or freelancers work on your content, share relevant prompts with them so outputs match your standards. Keep sensitive prompts, such as those involving internal data, for internal use only.
Prompts for different roles
Different team members need different prompts. Community managers need reply drafts and comment summaries; content creators need hooks, scripts and captions; analysts need report summaries and data questions; managers need briefs and strategy outlines. Tag prompts by role so each person finds what they need without wading through everything.
Placeholders and inputs
Use consistent placeholders such as square brackets for information people must add, and list the inputs at the top of each prompt. Clear placeholders stop people from running prompts with missing context, which is the most common reason for weak outputs.
Starter library size
Twenty to thirty well-tested prompts cover most marketing teams' daily work. Start small and grow based on demand.
Naming conventions
Name prompts by task and platform, such as caption, Instagram, launch, so searches work.
Example outputs
Store one or two good example outputs alongside each prompt. Examples show the expected quality and length, help new users judge whether their result is good enough and reveal when a tool update has changed how a prompt behaves.
Measuring the value
Track how often prompts are used and ask the team how much time they save. Compare the editing needed on drafts from library prompts with drafts from ad hoc prompts. Evidence of time saved helps justify the effort of maintaining the library and investing in better tools.
Prompts for analysis and reporting
Analysis prompts deserve the same care as writing prompts. Include instructions on how to treat the data, what to look for, how to present findings and a reminder to flag uncertainty. Ask for findings in plain English with the supporting numbers, so reviewers can check claims quickly.
A worked example
A five-person agency notices each team member prompts differently, producing inconsistent client work. It creates a shared library of 25 prompts with a brand voice block for each client, platform limits and example outputs. New starters use the library from day one, editing time falls and client feedback on voice consistency improves.
Common mistakes
- Prompts saved without context or inputs.
- No brand voice block.
- Outdated platform limits.
- Never testing or reviewing prompts.
- Libraries so complex nobody uses them.
Learning to work with AI
The Academy's AI tools module in the second level covers prompting, and the AI and automation at scale module in the leader level covers team workflows and governance. The free assessment shows where you would start.
