Hub › Lessons › Content planning, lesson 7
A 10-minute weekly content review with one AI prompt
Most content plans fail quietly. A post slips a week, a customer asks the same question twice, and nobody looks at the numbers until the month is over. This lesson gives you one short check to do every week. You copy three things into one prompt: your plan, last week's numbers and the questions customers asked. You get back what needs attention, what your numbers can and cannot tell you, and a plan for next week that fits the time you really have. It brings lessons 1 to 6 of this series together, but you can use it on its own.
What you will be able to do: in about 10 minutes a week, see which posts are late or undated, read last week's numbers without guessing at causes, decide where new customer questions belong, and leave with a plan for next week that is no bigger than you can make.
Step 1: Pick a fixed slot
- Same day, same time, every week. Friday afternoon works if you plan the week ahead on Monday. Put it in your calendar as a 15-minute block, so 10 minutes of work has room.
- Decide how many posts you can make next week before you start. Be honest: the number you actually made last week is a good guide. The prompt will not plan more than this.
Step 2: Copy in three things (5 minutes)
- Your plan. Every post that is not published yet, with its platform, stage and due date, or "no date". If you built the calendar in lesson 4, copy the rows from your "This week" view.
- Last week's numbers. For each post you published, copy the numbers the platform's insights show, using the platform's own words: views, impressions, saves, comments. Add anything that matters more than likes, such as a DM asking about price or an enquiry by email.
- Customer questions. Questions people asked you this week by message, email or in person, in their words, and how many people asked. These are the same questions lesson 1 turns into topics.
- Last week's change. If you tried one small change, write what it was and whether you did it.
Before you paste: remove customers' names, phone numbers and email addresses. The AI needs the question, not who asked it.
Step 3: The weekly review prompt
Fill in each line, add one line per post under the dashes, then paste it into your AI tool.
Help me do my weekly content review. Use only what I write below. Today's date: How many posts I can make next week: Posts on my plan (title, platform, stage, due date or "no date"): - Posts published this week (title, platform, date published, and the numbers I copied from the platform's insights): - Customer questions I got this week (in their words, and how many people asked): - The change I tried this week, and whether I did it: Return these parts: 1. Needs attention: posts on my plan that are late, due in the next 7 days but not Ready, or have no date. For each, one next action: finish it, move it to a new date, or drop it. 2. This week's numbers: a short table of the published posts with my numbers, unchanged. Then up to 2 observations. Only compare posts on the same platform. Say plainly when there are too few posts to tell why one did better. 3. This week's change: did I do it, and what my numbers can and cannot say about it. 4. Customer questions: for each question, say whether a post on my plan already answers it. If none does, suggest a new post idea with a working title. Mark any fact I would need to answer it as [ASK]. 5. Next week's plan: no more posts than I said I can make. For each: title, platform, due day, and the next step (write a brief, repurpose a published post, edit with my voice guide, or check the title against search results). Put late and nearly ready posts first. 6. One change to try next week, small enough to do in under 15 minutes, and how I will know I did it. 7. Not checked: one line on what this review cannot tell me. Rules: - Keep the review short: under 400 words, not counting the table in part 2. - Use only my notes and numbers. Do not add posts, numbers, percentages, benchmarks or facts I did not give. - Do not say that a change caused a result. One week of numbers only shows what happened. - Do not predict followers, reach, traffic or sales. - If I listed no published posts, write "Nothing published this week" in part 2 and do not judge the change in part 3. - If I listed no planned posts and no published posts, say there is nothing to review yet and stop after part 1.
Step 4: Decide, then update your calendar (3 minutes)
- Act on part 1 first. Give every late or undated post a new date, or drop it. A plan full of overdue posts stops being useful.
- Answer each [ASK] yourself before a question becomes a post. If you do not know the answer yet, the post waits.
- Copy next week's plan into your calendar with the due days, and put the one change somewhere you will see it.
- Read the review before you act on it. If it states a number or fact you did not give it, delete that part.
Worked example
We used the same made-up web designer as in lesson 5 and lesson 6, who builds simple websites for small cafes and home food businesses in Singapore. The business, posts, numbers and questions are illustrative. Two of the planned posts use titles from lesson 6.
Input (illustrative)
Today's date: Friday 30 October 2026 How many posts I can make next week: 2 Posts on my plan (title, platform, stage, due date or "no date"): - Before and after: a home bakery's new menu page | LinkedIn | Drafting | Tue 27 Oct - Should a cafe website menu show prices? Yes, and here's why | Blog | Drafting | Mon 2 Nov - Instagram or a website for your home bakery: when DM orders are enough | Blog | Ready | Wed 4 Nov - 3 photos every cafe menu page needs | Instagram | Idea | no date Posts published this week (title, platform, date published, and the numbers I copied from the platform's insights): - Put your opening hours at the top of your page | Instagram carousel | Mon 26 Oct | 412 views, 9 likes, 2 comments, 11 saves, 1 DM asking about price - Put your opening hours at the top of your page (same post as text) | LinkedIn | Tue 27 Oct | 180 impressions, 4 reactions, 0 comments - How long does a simple cafe website take? | Blog | Thu 29 Oct | 23 visits so far, 0 enquiries Customer questions I got this week (in their words, and how many people asked): - "Can I change the menu myself or do I have to pay you each time?" (2 people) - "Do you do online ordering?" (1 person) - "How much is a one-page website?" (1 person) The change I tried this week, and whether I did it: post the Instagram carousel on Monday morning instead of Thursday night. Yes, I did it.
Output (as returned)
Shown word for word from our test with Claude, without its title line; only the formatting was converted for this page.
1. Needs attention
- Before and after: a home bakery's new menu page (LinkedIn, Drafting): late since Tue 27 Oct. Next action: finish it.
- Should a cafe website menu show prices? (Blog, Drafting): due Mon 2 Nov and not Ready. Next action: move to a new date, because next week's two slots are taken (see part 5).
- 3 photos every cafe menu page needs (Instagram, Idea): no date. Next action: move it to a new date after next week.
2. This week's numbers
This week's numbers, as given. On a phone, scroll the table sideways.
| Post | Platform | Published | Your numbers |
|---|---|---|---|
| Put your opening hours at the top of your page | Instagram carousel | Mon 26 Oct | 412 views, 9 likes, 2 comments, 11 saves, 1 DM asking about price |
| Put your opening hours at the top of your page (text) | Tue 27 Oct | 180 impressions, 4 reactions, 0 comments | |
| How long does a simple cafe website take? | Blog | Thu 29 Oct | 23 visits so far, 0 enquiries |
- You have one post per platform, so there is no same-platform comparison. I can't say why any post did better than another.
- The Instagram carousel produced a DM asking about price. The blog post has no enquiries yet, but it has been live only since Thursday.
3. This week's change
You did it: the carousel went out Monday morning. Your numbers show what the carousel got. They can't show whether Monday morning beat Thursday night, because you gave me no Thursday-night post to compare. One week and one post can't show the timing made a difference.
4. Customer questions
- "Can I change the menu myself or do I have to pay you each time?" (2 people): no post answers it. Idea: "Can you update your cafe menu yourself? What to ask before hiring someone". [ASK] whether clients can edit their own menu, and what you charge for changes.
- "Do you do online ordering?" (1): no post answers it. The DM-orders post is about a different question. Idea: "Online ordering for a cafe: do you need it?" [ASK] whether you offer online ordering, and what it involves.
- "How much is a one-page website?" (1): no post answers it. Idea: "What a one-page cafe website costs". [ASK] your price or range, and what's included.
5. Next week's plan (2 posts)
- Before and after: a home bakery's new menu page, LinkedIn, due Mon 2 Nov. Next step: edit with my voice guide.
- Instagram or a website for your home bakery: when DM orders are enough, Blog, due Wed 4 Nov. Next step: check the title against search results.
The prices post doesn't fit in two slots, so it moves to a later date.
6. One change to try next week
Add one closing line to the "How long does a simple cafe website take?" post that invites readers to message you for a one-page website price (about 10 minutes). You'll know you did it when the line is live on the post.
7. Not checked
This review can't tell you why any post did what it did, or how these numbers compare with earlier weeks.
What we checked
- The numbers were not changed. A short script confirmed that every number in the table matches the input exactly, and that the only number in the output that is not in the input is "10" (minutes for the suggested change).
- No guessing at causes. The review did not claim the Monday posting time made a difference, and did not compare Instagram views with LinkedIn impressions, which count different things.
- The plan fits the time. Two posts were planned because the input said two. The late post and the Ready post came first, and the prices post was moved, not squeezed in.
- No invented answers. All three customer questions became ideas with [ASK] for the facts only the business knows, such as its price.
What we would do with this: the questions about menu changes and price are worth answering, because they came from real conversations and one matches the DM. But the web designer has to fill in each [ASK] first. Part 4 is a list of questions for the owner, not finished posts.
A quiet week
We also tested a week with nothing published, one undated idea and one question. Every run wrote "Nothing published this week", did not judge the change, and planned one post because the input said one. With an empty plan and nothing published, every run said there was nothing to review yet and stopped after part 1, as the prompt asks.
What we changed after testing
The prompt above is the third version. We ran each version on the same three inputs:
- Version 1 had no length limit. Both full reviews were over 700 words, which is too long to read in a short weekly slot. Version 2 asked for under 400 words.
- Version 2 shortened the review to about 500 words, but the table of numbers counted toward the limit. Version 3 leaves the table out of the count, so the numbers are never cut to save words.
- Version 3 still runs a little long. Our two full reviews were about 440 and 460 words without the table. That is a two- to three-minute read, so we kept it, but do not expect the AI to hit word limits exactly.
- What did not change: in all six full test runs, the numbers table matched the input, the plan had two posts with the late post and the Ready post, and no run claimed the change caused a result.
Where the other lessons fit
When the review points to a job, this is the lesson for it.
| When the review says | Use |
|---|---|
| New customer questions keep coming | Lesson 1: customer questions to content ideas |
| Next step: write a brief | Lesson 2: a one-page content brief |
| Next step: repurpose a published post | Lesson 3: one post into four short posts |
| Posts are late or have no date | Lesson 4: a Notion calendar with a "This week" view |
| Next step: edit with my voice guide | Lesson 5: a voice guide from your own posts |
| Next step: check the title against search results | Lesson 6: search intent check |
Watch out for
- One week is not a trend. A post that did well once may not do well again. Look for the same thing happening over several weeks before you change your plan.
- Platforms count differently. Views, impressions and visits are not the same measure, so compare posts on the same platform. Copy the platform's own words for each number.
- Likes are not customers. A message asking about price can tell you more about demand than many likes. Write those messages down.
- Keep private details out. Paste the question, not the customer's name or contact details.
- Do not fake a busy week. If you published nothing, say so. The prompt is built to handle a quiet week.
Next step
Set up the calendar this review reads from
Lesson 4 of this series builds a Notion calendar whose "This week" view lists late posts, posts due in the next 7 days and posts with no date, which is part 1 of this review.
Go to lesson 4Want your numbers in one place?
Our Notion Content Calendar template adds an engagement tracker where you copy each post's numbers, and a monthly review that adds them up. You still copy the numbers from each platform yourself. Everything in this lesson works without it.
See what's insideAll lessons
This is the last lesson of the content planning series. If you like a weekly routine, pair it with the weekly money ritual from the money admin series. The lessons page also has series on customer research and client onboarding.
See all lessonsHow we checked this lesson
- On 30 Oct 2026 (about 3:45 to 4:30 am Singapore time) we ran the prompt with Claude Code 2.1.312 in print mode on three illustrative inputs: a normal week, a quiet week with nothing published, and an empty plan. We ran version 1 twice per input, version 2 twice on the first two inputs and once on the empty plan, and version 3 twice on the normal week and once on each of the others. The output above is the first version 3 run on the normal week, shown as returned with formatting adjusted for this page.
- A separate script checked every output: the table numbers match the input, no new numbers apart from minutes in the suggested change, no words such as "caused", "benchmark" or "engagement rate", "Nothing published this week" for the quiet week, and the stop rule for the empty plan. We also read every output against the input.
- This lesson does not depend on any app feature, so there were no product facts to check on the day. The Content Calendar contents are taken from our own template's product page.
- We tested one assistant on three inputs. AI output varies by tool, model and input, so read the review each time.
- Last reviewed: 30 October 2026.