← Back to library
MarketingRoutine

Review which quote tweets worked and tune the scan

James Camp (@JamesonCamp) closes his X Article on quote tweets with a learning loop: a day after each one, log its numbers beside the score it got, then review weekly which signals predicted the winners and tighten the scan. Our version keeps rule changes as proposals you sign off on.

  • OutcomeEvery week, see which scan signals actually predicted your strongest quote tweets, with proposed rule changes you approve before the next morning scan uses them.
  • SetupMinutes
  • LevelStarter
  • CreditJames Camp (@JamesonCamp)
View original
Works with
Grok BotChatGPT agentClaudeOpenClaw

Published Oct 26, 2026Updated Oct 26, 2026AI dry run on sample input: passed Oct 26, 2026Source verified Oct 26, 2026

When should you use it?

  • You already run a daily quote-tweet shortlist and keep its scores
  • You want evidence before changing your threshold or quote ceiling
  • You post enough quote tweets each week to compare

Not for: Accounts with only a handful of quote tweets a week, where the numbers are mostly noise; Letting the agent quietly rewrite your scan rules on its own; Judging the content of your takes; this looks at which posts you chose and when

What do you need, and what do you get?

You need

  • The scan log from the daily shortlist: each candidate's score, quote count, age and metrics at scan time
  • Read access to your own post analytics on X: impressions, replies and reposts
  • A table where the results go, such as a sheet or Notion database
  • The review day and the current scan rules (threshold, ceiling, lists, terms)

You get

  • A row per quote tweet, filled about 24 hours after posting, with its numbers beside its scan score
  • A short weekly note: what predicted the best results, what predicted the worst, and how many posts the conclusion rests on
  • One or two proposed rule changes, each waiting for your yes before the scan uses it

How does it work, step by step?

  1. Link posts to scoresMake sure each quote tweet you publish can be matched to the candidate it came from, by the original post's link.
  2. Log after a dayAbout 24 hours after each one, record impressions, replies and reposts beside the score and the metrics the candidate had at scan time.
  3. Compare weeklyOnce a week, group results by signal: how old the original was, how many quotes it had, list versus search versus trending, and score band.
  4. Propose, do not applyHave the agent suggest at most two rule changes with the evidence behind each, such as lowering the quote ceiling. Approve or reject them yourself.
  5. Note the sampleAsk it to state how many posts each finding rests on, so a single viral outlier does not rewrite your rules.

The prompt

The original prompt by @JamesonCamp, word for word, with a link to where it was published. Below it is our adapted version with fill-in blanks, if you want it. Works in Grok Bot, ChatGPT agent, Claude and others. Values in lime are placeholders.

Original prompt by @JamesonCamp

Word for wordby @JamesonCamp
[24 hours] after each QT, log impressions, replies and
reposts next to the score it got in [my Notion table].
Every [Sunday], tell me which signals predicted my best
QTs and adjust the scan.

Exactly as published by @JamesonCamp in X Article: How You Can Get 5 Million Impressions a Month on X From Quote Tweets (Grok Bot + Drex Setup).

Our adapted version (optional)

Fill in the blanks (optional; the copied prompt uses your values)

Adapted prompt5 variables
Keep a results log for my quote tweets and review it weekly. Do not change my scan rules without my approval.
Results table: {results_table}. Scan log: {scan_log}. Review day: {review_day}. Current rules: {current_rules}.

1) About 24 hours after each quote tweet, add a row: link to my post, link to the original, impressions, replies, reposts, and the original's score, quote count and age at scan time.
2) On the review day, compare medians by signal: original's age when I quoted, existing quote count, where it was found (list, search, trending) and score band.
3) Tell me which signals went with my best and worst results, how many posts each finding rests on, and any outliers.
4) Propose at most two rule changes with the evidence. Wait for my yes before updating the scan.
5) If there are fewer than {min_posts} posts this week, say so and propose no changes.

What can go wrong?

  • One huge hit can dominate a week. Compare medians, not totals, and name outliers separately.
  • Impressions keep rising after a day. Use the same 24-hour mark every time so rows are comparable.
  • Changing several rules at once makes it impossible to tell which change helped. Adjust one or two per week.
  • Off-niche posts that worked because of a strong opinion can mislead the scan into widening your niche.

Our take

Most people who automate post-finding never close the loop, so their scan stays as good or as bad as the day they wrote it. Camp's second prompt is short and does the important thing: it ties each quote tweet's results back to the score that picked it, then asks what predicted the best ones. We kept his prompt word for word and changed one behavior. His version lets the agent adjust the scan; ours proposes at most two changes with the evidence and waits for you. A week of quote tweets is a small sample, and one viral post can make a bad rule look smart. We also ask for medians, a fixed 24-hour mark, and a count of how many posts back each claim. His example review, for instance, cut the quote ceiling after crowded posts underperformed, which is exactly the kind of single, testable change worth approving. In our dry run with ten fictional quote tweets, the agent filled each row at the 24-hour mark, found that the six posts quoted while the original still had fewer than ten quotes had a median of about 8,400 impressions, against about 1,900 for the four quoted after the original passed twenty, flagged one 61,000-impression outlier, proposed lowering the ceiling from 15 to 10, said the sample was too small to touch the score threshold, and changed nothing until approved.

Where the time goes back: Without a log, tuning the scan is guesswork. A weekly table lets you change one rule with evidence in a five-minute read.

FAQ

Does this need the daily shortlist playbook?
It works best with morning-quote-tweet-shortlist-you-write, because that saves each candidate's score and metrics. Without saved scores, the review can still compare timing and quote counts, but not the score itself.
Why not let the agent update the scan automatically?
Camp's prompt asks it to adjust the scan. We keep changes as proposals because a small or lopsided week can point the wrong way, and a rule you did not approve is hard to trace later.
Which signal usually matters most?
In Camp's own example report, early speed with few existing quotes did best and crowded posts did worst. Your data may differ, which is why the review exists.

Source and credit

The original prompt on this page is by @JamesonCamp, shown word for word with a link to where it was published. The summary, notes and adapted version are ours.

Original
X Article: How You Can Get 5 Million Impressions a Month on X From Quote Tweets (Grok Bot + Drex Setup)
Creator
James Camp (@JamesonCamp)
Published on
X
Original date
Oct 26, 2026
License
Not openly licensed; the original prompt is shown word for word with credit and a link to the original
Originally for
Grok Bot
AI dry run
Oct 26, 2026: passed. Dry-run tested by our AI test agent on sample input. Medians used; outlier named; sample size stated; change proposed, not applied.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails.
Verified
Oct 26, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.

Are you the creator? Claim or correct this page.

More in Marketing

All 97 Marketing playbooks →

Related playbooks

MarketingRoutine✓ AI dry run

Get a morning shortlist of posts worth a quote tweet

Each morning, receive up to five fresh posts from your niche that still have room for your view, each with its link, score, reason and one angle, while every word you post stays yours.

Credit: James Camp (@JamesonCamp)

Under an hour
MarketingPrompt✓ AI dry run

Answer a journalist's request using only facts you can back

For each media request, find out quickly whether you are a fit, then get a short answer-first draft built only from experience and data you confirmed, with unsupported lines flagged and the deadline checked.

Credit: Kourosh Ghaffari (@BuildingWaves)

Minutes
MarketingPrompt✓ AI dry run

Turn unlinked mentions of your brand into link requests

Get a short list of pages that already mention you without a link, each with the exact sentence, the page a link should point to, the public contact and a brief request draft you send yourself.

Credit: Kourosh Ghaffari (@BuildingWaves)

Under an hour