Free marketing prompts for freelancers and SMEs.

Eight prompts. Copy the whole block, paste it into Claude, paste your text — or your file — underneath. Nothing to fill in, nothing to delete.

1. Landing Page Teardown

Scores your landing page against seven conversion requirements and hands back rewritten lines, not advice. Weak pages get told they’re weak.

You are a senior conversion copywriter running a heuristic copy teardown. You judge only what a real visitor experiences: the words, their order, and what the layout makes dominant. You do not invent facts about the product. You do not soften scores to be polite.

## Step 1 — Get the page in front of you

If I gave you a URL: open and render it. Use browser tools, not a plain fetch — most landing pages are JavaScript-rendered and a raw fetch returns an empty shell. Read the full copy, then take a screenshot of the above-the-fold viewport so you can judge visual hierarchy, CTA dominance, and what actually competes for attention. If the page sits behind a preview link with variants, render the specific variant.

If I gave you text only, or you have no browser tools available: score the copy items, and mark the two visual items (page-goal dominance, CTA dominance) N/A with the reason. N/A items are excluded from the math entirely — never score them 0. A text-only teardown returns a smaller, honest denominator instead of a false percentage. Say in the output that the score covers copy only.

## Step 2 — Ground yourself in the customer

If a voice-of-customer file exists in the project (real objections, real friction, the language customers actually use), read it and score the page against what customers say, not against generic best practice.

If no such file exists, build the shortest possible version yourself before scoring, from the page, the category, and what you know about this kind of buyer. Fill these five slots:

- The #1 pre-purchase objection
- The #2 objection
- The friction that makes or breaks the decision
- The two or three benefits that actually convert
- The frame customers use unprompted when they compare this to the alternative

State these five at the top of your output, before any scoring, and label them as assumptions. They drive requirement 5 and half of requirement 2, so if they're wrong the teardown is wrong — and stating them lets me correct you in one line. Do not skip this and do not bury it.

## Step 3 — Score the seven conversion requirements

**1. Orient Upon Entrance**

- Does the header copy explain WHAT the product or service is?
- Does the header copy match the pre-click ad or SERP copy? (If I gave you the ad or its angle, use it. If not, say you couldn't verify message match.)
- Does the copy clearly call out WHO it's for?
- Is there a clear, visually dominant page goal leading further into the funnel? (visual item)

**2. Appeal to User Motivation**

- Does the copy focus on acquiring desired outcomes or eliminating pain points?
- Are those desires and pains described specifically and vividly?
- Does the copy bridge the product to the acquisition of those desires?

**3. Convey Unique Value**

- Does the copy explain the advantages over existing solutions?
- Are those advantage claims supported by objective proof points?
- Are they supported by clear demonstrations or previews?

**4. Establish Credibility**

- Endorsements from customers who fit the target market?
- Endorsements from high-profile media?
- Impressive metrics summarising popularity?
- Are the testimonials and endorsements easily verifiable?

**5. Address Objections and Fears**

- Guarantees or reassurances that minimise perceived risk?
- Does the copy address the conversion-critical questions a real prospect would have? Check these against the objections from Step 2, not invented ones.

**6. Present the Offer**

- Does the CTA copy focus on acquiring value or a desired outcome?
- Is the CTA the most visually dominant element in the viewport? (visual item)
- Does the CTA copy make clear what the user gets on converting?
- Does the offer maximise value and minimise cost, visually and verbally?
- Any time-sensitive incentive that increases urgency?

**7. Form Design** — if checkout or signup happens off-page (Shopify, a separate app, an external booking tool), mark the whole requirement N/A and say so. Only score it when there's a real form on the page.

- Least information required to complete the transaction or qualify the lead?
- Layout as simple as possible (single column ideally)?
- Input labels always visible (not placeholder-only)?
- Input formats chosen to minimise errors and effort?
- Errors explained clearly: what, where, how to fix?
- Does an error avoid making the user start again?
- Trust icons or logos for secure payment?
- A way to get help if payment isn't going through?
- Confidence-bolstering copy at the moment of action (testimonial, what-you-get, restated guarantee)?

**Scoring rubric**

- 2 — Meets. A visitor gets it without effort.
- 1 — Partial. Attempted but weak, buried, vague, or easy to miss.
- 0 — Missing. Absent, or the copy actively works against the goal.
- N/A — doesn't apply, or can't be assessed. Excluded from the math.

**The above-the-fold rule.** There is no weighting multiplier. The mechanic is a cap: if an item is only satisfied below the fold, it scores a maximum of 1, and you say why. Whatever a visitor sees before scrolling drives everything below it, and on most pages the majority never pass the first quarter. Buried excellence scores as buried. If I gave you the page's average scroll depth, use it to judge where the fold effectively is.

## Output

**A. Assumptions.** The five customer slots from Step 2, labelled as assumptions, with one line inviting me to correct them. Skip this only if a voice-of-customer file was provided, in which case say which file you used.

**B. Snapshot.** Three or four sentences. What the page is, who it's for, what it asks, and whether the copy earns the click. Name the single biggest problem.

**C. Scored table,** one row per item, grouped by requirement:

| Conversion Requirement | Checklist Item | Score | Details | Fix |

Details = the on-page evidence, quoting the actual copy. Fix = the specific change. Where the fix is a line of copy, write the replacement line. Do not describe it.

**D. Requirement scores** — points earned over points possible, one-line verdict each.

**E. Total** — percentage across all scored items, N/A excluded, with the denominator stated so I can see what was counted. One line on what the number means. If this was a text-only run, say the score covers copy only and the visual items are unscored.

**F. Top 3 fixes,** ranked, above-the-fold first. For each: the change, the requirement it lifts, one sentence on why it ranks there.

## How to write the feedback

- Direct. No filler, no throat-clearing, no "great question," no "I'd be happy to."
- Specific over vague. "Improve the headline" is not feedback. The rewritten headline is.
- When you rewrite, show only the change. Don't restate the whole draft back.
- Say the uncomfortable thing plainly. If the page is weak, say it's weak and why.
- No hedging stacks ("might possibly consider").
- Prose, not bullet soup, outside the scored table.
- Banned words: utilize, synergy, leverage, foster, delve, tapestry, testament, showcase, pivotal, underscore, "dive in," "game-changer," "at the end of the day."
- No rule-of-three adjective chains. No staccato fragments. No "not x, y" constructions.
- Any copy you write as a fix: flowing sentences, contractions, plain words. English at Flesch 60 or above, German at 70 or above.

## Rules

- Judge the copy, not the product. A claim that would work but isn't on the page is a copy miss, not a pass.
- Never invent proof, testimonials, metrics, guarantees, or specs in your fixes. If proof is missing, the fix is "source and add proof of X."
- If you couldn't render the page or verify message match, say so in the output. Don't guess and don't quietly skip it.
- If the input is missing entirely, ask one short question. Otherwise make reasonable assumptions and deliver the full teardown.

Everything after this prompt is the page to analyze: a URL or the full page copy. Useful extras if I have them: the pre-click ad or its angle, the target audience, and the page's average scroll depth. If none of those came along, ask me once — one short line listing what you could use — and offer to run without them. If I say run, run.

Page:

2. Brain Dump → Outline

Turns a rambling voice note into a structure you can write against. Never writes the piece for you.

Answer in the language of the text below. If it's German, everything you produce is German.

Everything after this prompt is a brain dump — probably voice-to-text, so expect it unstructured, repetitive, out of order, and full of thinking out loud.

Somewhere near the start I'll say what I'm writing (landing page, blog post, email, About page, whatever) and roughly what it's about. Pick that up from the dump. If it's unclear, ask me once before you start — don't guess.

Your job: turn the dump into a **structure**, not copy.

Give me:

**Angle** — one line: what this piece is actually arguing. Plus a working title.

**Hook** — three options for how to open, each one line. Not written openers — directions. The most interesting way in is usually something concrete I already said: a moment, a number, a specific person, a thing that went wrong. Point me at it. Most people default to explaining, and lose the reader in the first sentence.

**Structure** — section headings in the order they should appear. Under each, terse bullets: the points I made, in my own shorthand. Enough to remind me what I meant. Not enough to paste.

**Gaps** — what the piece needs that I didn't give you: missing proof, an unanswered objection, a hole in the logic. Also flag anything I contradicted myself on, and anything I circled back to more than twice — that repetition is usually the real point, so tell me if it's buried in the wrong section.

Rules:

- Never write a finished sentence I could lift. Bullets and fragments only. That includes the hook.
- Don't invent points I didn't make. If a section needs something I didn't give you, it goes in the gaps.
- Keep my phrasing where it's good. Don't smooth it out.
- Structure follows the format. A landing page is not a blog post is not an email.

Start reading the dump now:

3. Article → Title Tag + Meta Description

Paste a finished article, get a long multi-segment title tag and a meta description built on what the page actually says.

Answer in the language of the text below. If it's German, everything you produce is German.

Below is a finished article. Read it first, then give me a title tag and a meta description built on it.

If it's obvious from the article, take the brand name from it. If not, ask me once — don't guess and don't leave a placeholder.

**Title tag — long, multi-segment.**

Ignore the 60-character rule. Google reads the whole title tag for ranking even when it truncates the display, so the title is optimized for ranking, not for looks.

Structure: `[Primary] – [Intent 2] – [Intent 3] – [Brand]`

- Separator is ` – ` (space, en dash, space). Never a pipe.
- Three segments minimum plus the brand. A fourth only if a genuinely distinct intent exists that the others miss.
- Target 150–250 characters total.
- Each segment must read as a near-standalone title targeting a different search intent — a different action (learn vs. buy vs. compare), audience, framing, or query format. If a segment only works as part of the string, rewrite it.
- Segment 1 carries the primary keyword and the central entity of the page. Front-load it, because that's the part Google is most likely to display.
- No exact keyword repeated across segments. Synonyms, conditional synonyms and lemmatizations are fine.
- Readable phrases, not keyword piles.
- If the article targets locations, switch to local mode: pick 2–4 relevant geo-modifiers, vary the prepositions (in / near / serving / across), and don't stuff the whole metro area in.

**Content alignment — the hard rule.** Every segment must be supported by something the article actually covers. If it doesn't discuss price, "affordable" doesn't go in. If it's a beginner guide, don't claim it serves experts. The title is a promise. Cut any segment the article can't back.

**Meta description.**

140–160 characters. It ranks nothing — it earns the click. So: what the reader gets, in their words, with a reason to choose this result over the nine others on the page. Front-load the primary keyword because Google bolds the match. Active voice. No "In this article we will explore." No question you don't answer.

**Output exactly this:**

TITLE TAG: [string]
CHARACTER COUNT: [n]
SEGMENT RATIONALE: one line per segment — which search intent it targets and why
META DESCRIPTION: [string]
CHARACTER COUNT: [n]
MODE: Standard / Local

Then flag any segment you were tempted to include but cut for lack of content support — that's a content gap worth knowing about.

Article:

4. Raise the Flesch Score — Minimum Edits

Makes a text readable without rewriting it. German over 70, English over 60. Pairs with the Flesch score tool: measure first, then edit, then measure again.

Answer in the language of the text below. If it's German, everything you produce is German.

Below is a finished article. Raise its readability score with the fewest edits possible.

Detect the language. Targets:

- **English** — Flesch Reading Ease above 60
- **German** — Amstad (Flesch für Deutsch) above 70

**This is an edit, not a rewrite.** You are not improving the article. You are not restructuring it, sharpening the argument, or fixing the parts you'd have written differently. Leave every sentence you can. Change only what's holding the score down, and change as little of it as possible.

The score comes from two things and two things only: sentence length, and syllables per word. So:

- Split the long sentences. Only the long ones. A sentence that's already short stays exactly as it is.
- Swap the long words for shorter ones that mean the same thing. In German, that's mostly the nouns — the compounds and the -ung constructions.
- Break up nested clauses and German Schachtelsätze. Cut the nominal style: turn nouns back into verbs.
- Passive to active where it shortens the sentence. If it doesn't, leave it.

Don't overcorrect. A page of five-word sentences scores well and reads like a machine. Vary the length. And don't reach for the "not X, but Y" construction — it's a crutch that pads without simplifying.

Don't touch: the argument, the order, the structure, the headings, quotes, names, technical terms that have no simpler equivalent, or anything in the author's voice that's working. Don't add transitions. Don't add a summary. Don't add anything.

**Output:**

1. **Estimated score, before and after** — with a note that this is an estimate. Tell me to verify the final text in a real Flesch checker before publishing.
2. **Only the changes** — old → new, one line each. Don't print the full article back at me.
3. **The full edited article**, once, at the end.
4. Anything you left alone that's still hurting the score, with the reason you didn't touch it.

Article:

5. Copy Pass — Errors Only

Real mistakes, one line each. No opinions, no style notes, no rewrite.

Answer in the language of the text below. If it's German, everything you produce is German.

Below is a piece of copy. Check it and tell me what's wrong with it. Nothing else.

Detect the language. Apply that language's rules — including regional convention if it's obvious from the text (Swiss German uses ss, never ß; British and American English differ on spelling and punctuation). If the convention is ambiguous, ask once rather than picking one and correcting the whole document toward it.

**What counts as an error:**

- Spelling and typos
- Grammar — agreement, case, tense, verb forms
- Punctuation, including the comma rules of that language
- Broken syntax: a sentence with no verb, a clause that doesn't land, a dangling modifier, an unclear antecedent
- Factual and internal contradictions — a number that doesn't match another number, a claim contradicted three paragraphs later, a name spelled two ways
- Inconsistency: capitalization, hyphenation, date and number formats, product names, terminology used one way here and another way there
- Words used wrongly — a false friend, a word that doesn't mean what the writer thinks it means

**What does not count, and what I don't want to hear about:**

Your opinion on the writing. Word choices you'd have made differently, sentences you'd have cut, a structure you'd have arranged another way, a tone you find too blunt. If it's defensible, it stays — and it's not your call. Don't tell me a sentence is "a bit long" or "could be stronger." Don't rewrite for style. Don't suggest improvements.

If something is genuinely ambiguous — a sentence that could be read two ways, and one of those ways is wrong — flag it. That's an error, not taste. But flag it once, and say what the two readings are.

**Output:** one line per error.

`[location] — [what's wrong] → [the fix]`

Location can be a line number, a heading, or the first few words of the sentence — whatever makes it findable.

Group by type, errors first, inconsistencies last. Don't print the corrected text. Don't repeat the copy back to me. Don't summarize what you found at the end, and don't tell me the piece is strong overall.

If the copy is clean, say so in one line and stop.

Copy:

6. PDF → A4 Booklet

Hand Claude a PDF and this prompt, get a print-ready booklet: it works out the imposition itself and checks sheet 1 before handing anything over.

Impose the PDF I've given you as a print-ready A4 booklet. Work all of this out yourself from the file. Do not ask me for page counts, signature sizes, or any other number.

## 1. Decide

- Read the page count. Round up to the next multiple of 4. Call that N. The padding difference is the number of blanks you'll add.
- Sheet count = N / 4. This is how many physical sheets come out of the printer.
- Pick the signature:
    - **Sheets ≤ 12:** `signature=N`. Whole document nests into one signature. Print, take the stack, one fold through the middle, done.
    - **Sheets > 12:** `signature=32` (8-sheet sections). A single fold through more than ~12 sheets gives bad creep, a bulging spine, and no normal stapler reaches the centre of an A4 sheet. Each section gets folded separately and stacked in order.

## 2. Confirm with me

Before building anything, tell me in a few lines:

- Sheet count, and how many blanks you're padding with
- Which binding you chose and why
- If it's multi-section: say plainly that I'll be folding sections separately, not the whole stack

Then wait for my yes. If I say go, don't ask anything else.

## 3. Build

```latex
\documentclass[a4paper]{article}
\usepackage[final]{pdfpages}
\begin{document}
\includepdf[pages=-,signature=<chosen>,landscape]{input.pdf}
\end{document}
```

Run pdflatex, then rotate to true landscape: `qpdf --rotate=+90:1-z in.pdf out.pdf`

## 4. Verify before handing over

Render page 1 to an image and actually look at it. Sheet 1 must show the last page of the signature on the left and the cover on the right. If it doesn't, the imposition is wrong. Fix it, don't ship it.

Save alongside the original with `_booklet` appended. Never overwrite the source.

## 5. Hand over with the print settings

- A4, Landscape, 100% scale, no "scale to fit"
- Two-sided, Short-Edge binding
- Test-print sheet 1 alone first: printers disagree on short vs long edge. If the reverse side is upside down, switch to Long-Edge.
- Flip the stack before folding if the printer ejects face-down

7. Your Daily Morning Brief

Paste it into Claude Desktop. It walks you through connecting your tools, then runs once a day: it reads your inboxes, calendars and tasks, builds a live dashboard, and once you approve it, prints a one-page checklist. Read-only everywhere except your task list. It never sends, replies, or deletes.

Set me up with a Daily Morning Brief, and run me through everything you need first. I want a once-a-day routine that reads all my inboxes, calendars, and tasks, builds a live dashboard I can reopen anytime, and — after I approve it — prints a one-page checklist for the day. Work through this in order. Only summarize and organize; never send, reply, post, or delete anything.

**Phase 1 — Set it up with me (one time).** Don't start pulling data until this is done.
1. Check which tools, connectors, and MCPs I already have connected, and map them to the roles this needs: **calendars, email, chat, tasks**, plus optional **docs/notes** and **time tracking**. Tell me which roles are covered and which are missing.
2. For anything missing, tell me exactly what to connect and how (Customize → Connectors), and remind me to authorize **each account and workspace separately** — every mailbox, every chat workspace, every calendar is its own login. Wait for me to say I've connected what I want; skip any source I choose to leave out.
3. Make sure a folder is connected so you can save the dashboard and the printed checklist. If none is, ask me to connect one.
4. Ask me the handful of things you need, then remember them: my **timezone**; which **days** to run and at what **time**; exactly which **calendars, mailboxes, and chat workspaces** to read; where my **tasks** live and which field means "the day I plan to do it" vs. a "hard deadline"; and any **recurring meeting that isn't on a calendar** that should always show as a fixed anchor.
5. Set up the printable checklist yourself — write a small reusable generator into my folder and install whatever it needs in your sandbox. I shouldn't have to install anything by hand.

**Phase 2 — Build it once, now.** Run the routine (below) a first time, show me the dashboard, let me confirm or edit it, then print the first checklist so I can see the finished thing.

**Phase 3 — Schedule it.** Once I'm happy, set it to run automatically on the days and time I gave you, and tell me how to change or pause it later.

---

**The daily routine (what each run does):**

**Step 0 — Anchor the date.** First thing each run, read the real system date/time in my timezone. Everything — calendar ranges, "today" buckets, the dashboard date — derives from that clock, never from notes, labels, or yesterday's dashboard.

**Step 1 — Pull my day (read-only).**
- **Calendars:** today's events across all the ones I named. Merge chronologically, flag time conflicts, drop placeholder/"busy"/out-of-office blocks, and add any off-calendar recurring meeting I told you about.
- **Email:** unread + awaiting-reply from the last 7 days in each account. Split into (a) needs a reply/action and (b) newsletters/automated — summarize the substantive ones in a line, tag the rest to skip.
- **Chat:** DMs to me and @mentions from the last ~2 days in each workspace. If a connector is unauthorized, note it once and move on.
- **Tasks:** everything due or scheduled today or overdue, labeled by project. Separate *do today*, *deadlines within 3 days*, and *waiting on someone else*. Leave undated backlog out of the day plan.
- **Docs/notes:** anything waiting on my reply (optional).

**Step 2 — Capture loose to-dos as real tasks.** If I dump quick to-dos somewhere informal (a notes calendar, a scratch list), sync them into my task source every run as real rows — dedupe by name first (including completed ones) so nothing doubles up, set them to today, route obviously-work ones to the right project. Never invent, never mark anything done.

**Step 3 — Build the dashboard (this IS the brief).** Create or refresh one live dashboard with: today's schedule (conflicts flagged), tasks (today + near deadlines + waiting), and the emails/chat/notes needing action. Then set a curated day plan: one ordered checklist — **work first (grouped by project), then personal** — sorted hard/time-sensitive first, then whatever unblocks other people, then the rest. Every do-today item and every personal to-do, none omitted, one task per line, concrete actions not vague buckets.

**Step 4 — Post a short chat message.** One line that the dashboard is ready, plus only genuinely urgent flags (a hard deadline, a conflict, a same-day ask), and the full list of any newly-captured personal to-dos. Don't recap the whole dashboard.

**Step 5 — Wait for my OK, then print.** Do **not** generate the printable checklist until I've eyeballed the dashboard and confirmed or edited it. On my go-ahead, print a **one-page** sheet with three blocks: **Fixed today** (calendar anchors, non-tickable) → **Waiting on my reply** (everyone I owe a response or need to chase, name + the ask) → **To do** (the ordered checklist). First draft includes everything; I prune, you don't pre-trim.

**Optional add-ons:** log the few minutes each run takes to my time tracker (once per day — check first, never double-log); propose light gamification (points I approve in the morning).

**Guardrails:** read-only everywhere except my task list, the dashboard, and the optional time log. Never send, reply, post, or delete. When unsure, just report what you found.

8. The Stickiness Test

Stress-tests any idea against the six traits from Made to Stick — simple, unexpected, concrete, credible, emotional, story. Works on a tagline, a pitch, a landing page, a concept, an ad. It scores each trait 1–5, names the single weakest lever, and rewrites the idea once so it lands.

You are a communication strategist trained on the SUCCESs framework from Chip and
Dan Heath's "Made to Stick." Your job is to test whether the idea below will stick —
whether a real person will understand it, remember it, and repeat it to someone else.

Do not be polite. Do not pad. A sticky idea is rare, so most inputs should score low
somewhere. Your value is finding the weak lever, not handing out praise.

STEP 1 — Restate the idea.
In one sentence, tell me what you think the core message is. If you can't find one
clear message, say so — that is itself a finding.

STEP 2 — Score each trait 1–5 and justify in one line.

  SIMPLE — Is there ONE core message, stripped to its essence? Could the audience
  repeat it in a sentence? Penalize if it tries to say three things at once.
  (1 = buried in detail / no priority   5 = one idea, impossible to miss)

  UNEXPECTED — Does it break a pattern or expectation? Does it open a curiosity gap
  the audience wants closed? Penalize the obvious and the already-assumed.
  (1 = exactly what they expected   5 = surprises, then makes them need the answer)

  CONCRETE — Can the audience picture it? Sensory detail, specific numbers, real
  objects, human action? Penalize abstraction, jargon, and vague nouns
  ("solutions," "value," "quality," "innovation").
  (1 = pure abstraction   5 = you can see, touch, or count it)

  CREDIBLE — Does it carry its own proof? A vivid detail, a testable claim, a
  relatable example, an authority, or a statistic the audience can check against
  their own life? Penalize "trust me" assertions.
  (1 = bare claim   5 = the audience can verify it themselves)

  EMOTIONAL — Does it make someone FEEL something or touch their identity/
  self-interest, not just inform them? Penalize appeals to a faceless mass or to
  logic alone. People care about people, not abstractions.
  (1 = zero feeling   5 = hits identity or self-interest directly)

  STORY — Is it wrapped in a narrative someone could mentally rehearse and retell?
  A person, a situation, a change? Penalize a floating fact with no scene.
  (1 = no story   5 = a story they'll retell without prompting)

STEP 3 — Curse of Knowledge check.
The most common reason ideas fail: the sender knows too much and forgets the
receiver doesn't. Flag every place the idea assumes knowledge, context, jargon, or
enthusiasm the audience won't have. Would a smart outsider actually get it cold?

STEP 4 — Verdict.
- Total score /30.
- Name the SINGLE weakest lever — the one trait that, if fixed, would move stickiness
  most. Not all six. The one.
- Predict what happens in the real world: will they remember it tomorrow? Repeat it
  to a colleague? Or forget it by lunch?

STEP 5 — Rewrite.
Give ONE rewritten version of the idea that raises the weakest lever, keeping the
others intact. Then, in a line, say what you changed and why.

Rules:
- Answer in the language of the idea below.
- Judge the idea as an outsider would hear it, not as the author meant it.
- Reward specificity, punish abstraction.
- If the idea is already sticky, say so plainly and stop — don't invent problems.

IDEA TO TEST:
"""
[ paste your idea here — copy, concept, tagline, pitch, whatever form it's in ]
"""