What can AI actually tell me about my social media? Far more than a summary of likes. Once your AI assistant is connected to your ZoomSphere workspace through the MCP, it reads your real posts, analytics, calendar and approvals, and answers the questions you would otherwise spend an afternoon in spreadsheets on: why reach dropped, which topic you have worn out, which posts deserve a second life, where client approvals get stuck. Here are 20 of those questions, grouped by what they help you decide, each with a prompt you can copy. Six of them come with a real example of what the answer looks like in Claude.

{{form-component}} 

Before you ask: one connection, three habits

Connect ZoomSphere to your AI assistant once. In Claude it is one click from the connector directory, in ChatGPT and other clients it is a custom connector; both paths are on our MCP page. From then on the assistant can read the workspaces your login can see.

Then three habits make the difference between a vague summary and an answer you can act on:

• Name the period. "September" or "the last 90 days", never "recently". The AI pulls exactly that window and compares it with the one before.
• Give it a benchmark from your own history. Your median, last month, the same month last year. A number with nothing to compare it to is decoration.
• Ask for an action. End every question with "and tell me what to do about it". That turns analysis into a decision.

Performance: what actually happened

1. What happened this month, in three sentences?

The question every client email starts with. Asking for exactly three sentences forces the AI to pick what matters instead of listing every metric it can find.

2. Which posts beat our own average, and what do they have in common?

Top posts lists are easy. The useful part is the pattern behind them: format, length, topic, time, whether a person was on camera.

3. Why did reach drop?

The question clients ask most nervously. A good answer separates what you did (fewer posts, a different format mix) from what happened to you, and checks which audience disappeared.

4. Which channel is worth the effort, and which is not?

Every channel costs time. This question puts the effort (posts, formats, production) next to the result, so the conversation about dropping or doubling a channel starts from data.

Content: what to make next

5. Which topic have we used up?

Recurring formats feel safe and quietly die. This question finds the themes whose performance drops with every repeat, before your client notices it first.

6. What worked last year that we have not posted this quarter?

Good ideas get forgotten when the team or the season changes. The AI remembers everything you published.

7. Which content pillar earns its slots?

Pillars get equal space in the plan and very unequal results. Comparing the share of posts with the share of results shows which pillar deserves more of the calendar.

8. Which format is winning this month?

Reels, carousels, single images, text posts: the balance shifts every few months. This question catches the shift while it is still useful.

{{cta-component}} 

Repurposing: get more from what already worked

9. Which three posts from this year should we repurpose, and how?

The cheapest good content is the content you already proved. The trick is choosing by the right signal: saves and reach to new people, not raw likes.

10. Which posts had lots of saves but little reach?

A high save rate with low reach means the content was useful and the algorithm simply did not show it. These are the safest posts to boost or post again.

11. Turn our best LinkedIn post into an Instagram carousel

Here the AI stops analysing and starts working. It reads the original, rewrites it for the new format, and leaves the result in the calendar for approval.

12. Which evergreen posts can we republish?

Some posts are as true today as the day they went out, and most of your current followers never saw them. The AI can find them and check they have not been reposted already.

Audience and timing

13. When should we post next month?

Generic best-time-to-post articles describe someone else's audience. This answer is built from yours.

14. Are we reaching new people, or the same fans?

Growth depends on reach outside your followers. This question shows which channel actually brings new people in.

15. Who actually follows us on LinkedIn?

For B2B clients, the question is not how many followers, but whether they can sign a contract.

16. What does our audience save, and what do they share?

Saves and shares are different signals. Saves mean "useful to me", shares mean "says something about me". Knowing which posts get which tells you what to make for each goal.

Workflow, approvals and the client

17. Where are our approvals getting stuck?

Analytics tools look at results. ZoomSphere also knows how the work moved: who approved what, how long it waited and what came back for rework. That makes questions like this one possible.

18. What is planned for next week, and where are the gaps?

The Monday morning check, done in one sentence: empty days, missing visuals, posts still waiting for approval.

19. Which client needs attention this month?

For agencies with many workspaces, the hardest part is noticing the account that is quietly slipping. One question across all clients catches it early.

20. Prepare me for tomorrow's client meeting

The question that saves the most stress. The AI anticipates what the client will ask, prepares the answers from data, and flags what you should mention before they do.

What your AI cannot answer (yet)

The answers are only as good as the data behind them, so it helps to know the edges:

• Exact algorithm causes. The AI can show what changed in your content and audience, but nobody outside the platforms knows precisely why a single post was or was not shown.
• Data that is not in ZoomSphere. Sales, website conversions or CRM data need their own connection. Connect them next to ZoomSphere and you can ask questions across both.
• Channels that are not connected. If a network is missing from your workspace, ask the AI to say so rather than leave it out silently. All the prompts above work best with that instruction.
• Taste. The AI can tell you which post worked and suggest the next one. Whether it fits the brand is still your call, which is exactly why its drafts wait for approval.

Turn the best questions into a routine

Once a question proves useful, stop asking it by hand. Questions 1, 18 and 19 make excellent scheduled tasks: a summary every first Monday, a gap check every Friday, a client health check once a month. Our guide to putting your social media on autopilot with scheduled AI tasks shows how to set them up.

When the answer needs to become a slide, our article on report charts worth sending has the prompts that turn these numbers into a client deck. And more than 30 ready prompts live in the ZoomSphere MCP prompt library.

Frequently asked questions

Which AI assistants can answer these questions?

Any assistant that supports MCP connectors: Claude, ChatGPT and others. Connect ZoomSphere once and the same prompts work everywhere.

Does the AI see all my clients' data?

It sees the workspaces your ZoomSphere login can access, through the MCP connection under your own account. If you only want it to look at one client, name the workspace in the prompt.

Are the example answers real?

They come from the ZoomSphere demo workspace with fictional clients, so the numbers are illustrative. The questions, prompts and the shape of the answers are exactly what you get on your own data.

Can the AI publish posts based on its answers?

It can create drafts, and we recommend keeping it that way. Posts land as Private Drafts in your calendar and go through your normal approval before anything is published.

How far back can the AI look?

As far back as your analytics in ZoomSphere go. For comparisons, ask for a specific window, such as the last 90 days or the same month last year, so the answer stays precise.

Do I need to know which metrics to ask for?

No. Ask the question in plain language. The AI chooses the metrics from what ZoomSphere offers and tells you which ones it used.

{{form-component}}