Content generation

VOICE GUIDE
Brand: [brand and product, e.g. invoicing for freelancers]
Who we write for: [audience, e.g. solo designers, 1-3 years in]
We sound: [3-4 traits as "X, not Y", e.g. warm, not cute]
Tone sliders, 1-5, on NN/g's four dimensions:
  formal [score, e.g. 2]      funny [score, e.g. 2]
  irreverent [score, e.g. 1]  enthusiastic [score, e.g. 3]
Reading level: [grade, e.g. 8; sentences under 18 words]
Say: [preferred terms, e.g. "clients", not "customers"]
Never say: [banned words, e.g. leverage, synergy, delve]
Format: [rules, e.g. sentence-case headings, no emoji]
Point of view: we = [who, e.g. the company]; you = the reader
Claims: never [banned claims, e.g. guarantees, unsourced %]
On-voice samples: [paste 2-3 passages]
Off-voice sample: [paste 1 passage, with why, e.g. too salesy]

| Time  | Visual (A/B-roll)       | On-screen text  | Voiceover             |
| 0-2s  | face to camera, close   | "97 days late"  | "My biggest client    |
|       |                         |                 |  paid 97 days late."  |
| 2-6s  | screen: overdue invoice | "Here's why"    | "It was my fault..."  |

<context>
I need one pillar article and the social, email and video
pieces cut from it. I'm not a content strategist.
Topic: [topic, e.g. how to track billable hours]
Business: [one line, e.g. a time-tracking app for freelancers]
</context>
 
<task>
Interview me the way a content lead briefs a writer before
commissioning a pillar article and its repurposed pieces.
Don't outline or draft anything yet.
</task>
 
<requirements>
- At most 10 questions, grouped: audience and search intent;
  angle and proof; channels and CTA; voice and constraints.
- For each: why the answer changes the piece, and 2-3
  typical answers I can pick from or adapt.
- Use the terms search intent, funnel stage, E-E-A-T and CTA,
  each defined in one line the first time.
</requirements>
 
<output_format>
Numbered questions under the four group headings, each
followed by "Why:" and "Typical answers:".
</output_format>
 
<process>
Ask all the questions in one message, then stop. When I
answer, write the brief back as bullets and list anything
still missing.
</process>

<materials>
[paste the brief from stage 1]
[paste the top 10 titles and URLs for the keyword]
</materials>
 
<task>
Propose 3 materially different approaches to the pillar
article, each with the repurposing plan it implies, and
recommend one.
</task>
 
<requirements>
- Approaches differ in format or angle, not wording: e.g. a
  step-by-step how-to, a data-led report on our survey, a
  contrarian argument.
- For each: a one-paragraph summary; the search intent and
  format it matches; its information gain over the top
  results; the proof it needs from me; effort; main risk;
  which channel pieces it feeds best.
- Recommend one for my goal and say what would change your
  mind.
</requirements>
 
<output_format>
A table: Approach | Intent and format | Information gain |
Proof needed | Effort | Risk | Best derivatives. Then the
recommendation in 3 sentences.
</output_format>
 
<process>
Stop after the recommendation and wait for my choice. Don't
outline or draft.
</process>

<materials>
Brief and chosen approach: [paste from stages 1-2]
Sources: [paste survey results, transcripts, product docs]
Keyword data: [paste from your keyword tool]
Voice guide: [paste]
</materials>
 
<context>
Readers: [audience, e.g. freelance designers who undercharge]
Primary keyword: [keyword, e.g. how to track billable hours]
Intent: [intent: informational | commercial | transactional]
Funnel stage: [stage: TOFU | MOFU | BOFU]
Channels for the pieces: [channels, e.g. LinkedIn, X, email]
</context>
 
<examples>
[paste 1-2 of our pieces that performed well, with why]
[paste 1 piece we disliked, with why, e.g. generic intro]
</examples>
 
<task>
Write the pillar article and its repurposed channel pieces.
The article must win the keyword by adding what the top
results lack; every piece leads to one CTA:
[CTA, e.g. start a free trial].
</task>
 
<requirements>
- SEO package: 5 title tags (50-60 characters, keyword near
  the front), H1, slug, meta description (about 150
  characters).
- Article: an intro that answers the query in 2-3 sentences;
  H2/H3s that follow searchers' sub-questions; proof where
  claims are made; internal links with descriptive anchor
  text to [paste URLs]; one CTA.
- E-E-A-T: mark each place that needs the author's first-hand
  experience [AUTHOR TO FILL].
- Pieces, each written from the article (not from another
  piece) with its own hook and CTA:
  [pieces, e.g. 1 LinkedIn post, a 6-post thread, 1 email].
- CTA blocks follow [structure: AIDA | PAS | a plain ask].
- Follow the voice guide; none of the AI tells listed in
  this sheet.
- No search volumes, platform limits or trends from memory.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Intent match: format and depth fit what the searcher wants
  to do.
- Information gain: at least 3 points the top results lack,
  each from my materials.
- Accuracy: every number, name and quote traces to my
  materials or carries [NEEDS SOURCE].
- Platform fit: each piece has its channel's native shape,
  length and hook, and stands alone.
- One CTA per piece, the same next step everywhere.
</quality_bar>
 
<output_format>
Markdown: the SEO package as a table; the article; each
piece under its own heading (video scripts as a table: Time |
Visual | On-screen text | Voiceover); a claims list:
claim | source in my materials, or UNSOURCED.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. Write the outline first
and stop for my OK. Then write the article section by
section, then the pieces. End with a list of the
assumptions you made.
</process>

<materials>
Brief: [paste]
Draft article and pieces: [paste]
Sources the draft was meant to use: [paste]
</materials>
 
<context>
This is a fresh chat. The draft was written by a model in
another chat; I need an editor's verdict before publishing.
</context>
 
<task>
Review the article and its pieces against the rubric below
and rank the fixes. Don't rewrite anything.
</task>
 
<requirements>
- Score 1-5, with one line of quoted evidence each: intent
  match; originality (information gain); accuracy; voice
  match with the guide; platform fit; one CTA; SEO hygiene
  (keyword placement, descriptive anchors, real author);
  AI tells.
- For every score under 4, quote the passage and say what a
  fix must achieve, not the new wording.
- List every factual claim with its source in my materials,
  or UNSOURCED.
</requirements>
 
<output_format>
Score table (Criterion | Score | Evidence); at most 7 fixes,
worst first; the claims list.
</output_format>
 
<process>
End with a list of the assumptions you made and any
criterion you couldn't judge from what I pasted.
</process>

Write a voice guide from these samples: 3-4 traits as "X, not
Y", sentence length, words we use and never use, how we open
and close. Quote the sample behind each rule.
Check before answering: every rule is visible in a sample.
[paste 3-10 pieces, each labeled with channel and result]

<materials>
[paste 3-10 top pieces, each labeled with channel and result]
</materials>
 
<context>
We publish on [channels, e.g. a blog, LinkedIn, a newsletter]
for [audience, e.g. freelance designers]. Several people and
a chat assistant write for us; drafts drift toward a generic
assistant voice.
</context>
 
<examples>
[paste 1 off-voice piece, with why, e.g. stiff and salesy]
</examples>
 
<task>
Write a voice guide that any writer or chat assistant can
follow to sound like our best pieces. It will be pasted into
every content prompt.
</task>
 
<requirements>
- Personality: 3-4 traits, each as "X, not Y".
- Tone by situation: [situations, e.g. launch, outage, billing]
- Tone sliders 1-5 on NN/g's four dimensions: formal-casual,
  serious-funny, respectful-irreverent,
  matter-of-fact-enthusiastic.
- Mechanics: sentence and paragraph length, reading level,
  heading case, emoji, numbers, Oxford comma.
- Words we use; words we never use; how we open and close.
- Two short on-voice passages on a new topic, and one
  off-voice passage with the reason.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
- Under 400 words.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every rule quotes the sample it came from.
- Each trait is specific enough to reject a draft ("warm, not
  cute"), not an adjective any brand could claim.
- Nothing contradicts a sample that performed well.
- A new writer could apply it without asking anything.
</quality_bar>
 
<output_format>
The guide in this sheet's VOICE GUIDE layout, plain text.
Then a table: rule | evidence quoted from a sample.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made, including rules drawn from only one
sample.
</process>

Group this keyword export into clusters one page could rank
for. Per cluster: working title, intent, primary keyword,
format, priority. Flag cannibalization. Use only my data.
Check before answering: no keyword sits in two clusters.
[paste CSV: keyword, volume, difficulty]

<materials>
Keyword export: [paste CSV: keyword, volume, difficulty]
Existing pages: [paste URLs with their target keywords]
</materials>
 
<context>
Product: [product, e.g. a time-tracking app for freelancers]
Capacity: [capacity, e.g. 4 articles a month]
Site: [site: new | established in a niche | a strong domain]
</context>
 
<task>
Turn the export into a keyword cluster map: which pages to
write, in what order, and the pillar page they hang from.
</task>
 
<requirements>
- One cluster = queries one page could rank for.
- Classify each cluster's search intent (informational,
  commercial, transactional, navigational) and give the
  format that usually matches.
- Priority high, medium or low from volume, difficulty and
  fit with the product; say which drove each call. Favor
  long-tail clusters a site like ours can win.
- Propose one pillar page and the clusters that link to it.
- Flag pages that would cannibalize each other or an
  existing URL.
- No volumes or difficulty from memory. Never invent
  numbers, names, quotes or sources: write [NEEDS SOURCE]
  where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every keyword appears in exactly one cluster.
- Each cluster has one intent; mixed clusters are split.
- Priorities follow from the pasted numbers.
- No two planned pages share a primary keyword.
</quality_bar>
 
<output_format>
A table: Cluster | Working title | Intent | Format | Primary
keyword | Secondary keywords | Total volume | Difficulty |
Priority. Then the pillar map as an indented outline, then
the cannibalization flags.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

Build a [month, e.g. March] content calendar: Date | Channel |
Pillar | Format | Hook | Funnel stage | CTA | Source asset.
Pillars: [pillars, e.g. pricing, workflow, clients]
Check before answering: every source asset feeds 3+ channels.
[paste channels, cadence, source assets and key dates]

<materials>
Source assets: [paste articles, webinars, surveys, launches]
Last quarter by channel: [paste results, if you have them]
</materials>
 
<context>
Pillars: [3-5 pillars, e.g. pricing, workflow, clients]
Cadence: [cadence, e.g. blog 2/week, LinkedIn 3/week]
Key dates: [dates, e.g. a launch on the 14th]
Capacity: [capacity, e.g. one writer, 10 hours a week]
Month type: [mix: steady teaching | launch month | event month]
</context>
 
<task>
Plan [month, e.g. March] as a calendar the team can run
without another meeting.
</task>
 
<requirements>
- Every source asset feeds at least 3 channels.
- No two promotional posts on consecutive days; each pillar
  appears at least [count, e.g. 3] times.
- Mostly TOFU, with MOFU and BOFU pieces around key dates.
- A cadence the capacity can sustain for a quarter.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The hours fit the team's capacity.
- Pillars and formats rotate; no week is all promotion.
- Every row has one CTA and a source asset.
- Each repurposed piece comes after the piece it's cut from.
</quality_bar>
 
<output_format>
A table: Date | Channel | Pillar | Format | Working hook |
Funnel stage | CTA | Source asset | Owner | Status. Then what
I must supply (data, quotes, footage) and by when.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

Write this week's newsletter for
[audience, e.g. freelance designers] from my notes: one main
story, 2-3 labeled sections, one CTA, 5 subject + preheader
pairs. Check: each preheader adds to its subject.
[paste notes and links]

<materials>
Main story: [paste notes, data or the source article]
Secondary items: [paste 2-3 links or notes]
Voice guide: [paste]
</materials>
 
<context>
[newsletter, e.g. a weekly email to 8,000 freelancers]
Format: [format: one-story essay | curated links | product news]
We judge issues by clicks and replies, not opens.
</context>
 
<examples>
[paste 1-2 past issues that got clicks or replies, with why]
[paste 1 issue that flopped, with why, e.g. three CTAs]
</examples>
 
<task>
Write issue [number, e.g. 42] so that readers
[CTA, e.g. register for the pricing workshop].
</task>
 
<requirements>
- Subject lines about 30-50 characters; each preheader adds
  information and never repeats the subject.
- A one-sentence opening; the main story in about
  [length, e.g. 300] words; labeled short sections; one CTA
  button text.
- The CTA block follows [structure: AIDA | PAS | a plain ask].
- No "Hope you're well", no "In this issue", no sign-off
  clichés.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Each subject + preheader pair makes one promise the issue
  keeps.
- A skimmer gets the main story from labels and first
  sentences.
- Exactly one CTA, matching the goal.
- It sounds like the issues that worked, not an assistant.
</quality_bar>
 
<output_format>
1. Five subject + preheader pairs, each labeled with the
promise or emotion it uses. 2. The issue in Markdown.
3. A plain-text version.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

Write a [length, e.g. 30]-second vertical video script about
[idea, e.g. why invoices get paid late] as a table: Time |
Visual | On-screen text | Voiceover. Hook in 0-3 s, a visual
change every 4 s. Check: the ending pays off the hook.
[paste source material]

<materials>
[paste the source: a post, article section or transcript]
</materials>
 
<context>
Platform: [platform: TikTok | Instagram Reels | YouTube Shorts]
9:16, captions on. Audience: [audience, e.g. freelancers]
Goal: [goal: follows | clicks | sign-ups]
Filmed by: [crew, e.g. me on a phone, one take per beat]
</context>
 
<examples>
[paste 1-2 scripts that held retention, with why]
[paste 1 that lost viewers early, with why, e.g. greeting]
</examples>
 
<task>
Write a [length, e.g. 30]-second script built on the single
most surprising point in the source.
</task>
 
<requirements>
- 0-3 s: a hook that states the payoff or a tension; no
  greeting, no "Hey guys".
- A visual change at least every [seconds, e.g. 4] seconds.
- On-screen text: at most 6 words per beat, in the safe zone.
- Ending: [ending: a CTA | a line that loops to the opening]
- Only claims from the source. Never invent numbers, names,
  quotes or sources: write [NEEDS SOURCE] where support is
  missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The first 3 seconds earn the next 3, and the end pays off
  the hook.
- Read aloud, the voiceover fits the length (roughly 2-3
  words a second).
- Every beat has a visual a phone can film.
- On-screen text adds to the voiceover instead of repeating
  it.
</quality_bar>
 
<output_format>
A table: Time | Visual (A-roll or B-roll) | On-screen text |
Voiceover. Then 3 alternative hooks, a caption, a cover-frame
idea and a B-roll shot list.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

Propose 4 thumbnail concepts for
"[title, e.g. My client paid 97 days late]": focal point,
emotion, 2-4 overlay words, layout, and an image prompt each
(no text in the image). Check: readable at phone size; it
promises only what the content delivers. [paste a summary]

<materials>
[paste a summary of the video or article]
</materials>
 
<context>
For: [placement: YouTube thumbnail | blog header | link preview]
Aspect ratio: [ratio: 16:9 | 1.91:1 | 4:5 | 9:16]
Brand colors: [hex values, e.g. #0F766E, #F8FAFC]
Competitors look like: [style, e.g. shocked faces, arrows]
</context>
 
<examples>
[paste or describe 1-2 thumbnails you like, with why]
[paste or describe 1 you dislike, with why, e.g. clickbait]
</examples>
 
<task>
Propose 4 thumbnail concepts and an image-generator prompt
for each, so I can generate the image and add the words in a
design tool.
</task>
 
<requirements>
- Per concept: one focal point, the emotion or tension, 2-4
  overlay words (or none), where text and subject sit.
- Image prompts: subject, composition with space for text,
  [style: editorial photo | flat illustration | 3D render],
  lighting, palette, aspect ratio, what to avoid (stock
  handshakes, glowing brains, robots).
- No text inside generated images; no real person's likeness
  without permission.
- The thumbnail promises nothing the content doesn't deliver.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- At phone size, the focal point and words read in a glance.
- Title and thumbnail work as one unit and don't repeat each
  other.
- Strong contrast between subject, text area and background.
- The promise matches the content.
</quality_bar>
 
<output_format>
A table: Concept | Focal point | Emotion | Overlay words |
Layout. Then the 4 image prompts, one block each.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

This article lost [drop, e.g. 40%] of its traffic. Compare it
with the current top results and list what to update, add,
merge or cut, and which facts to re-verify. Don't rewrite.
Check: keep the URL and the parts that still rank.
[paste the article, the top results and any new facts]

<materials>
Article: [paste, with its publish date]
Top results now: [paste titles and URLs for its keyword]
New information since publication: [paste]
Analytics: [paste monthly clicks and positions]
</materials>
 
<context>
Keyword: [keyword, e.g. freelance invoice template]
Traffic fell [drop, e.g. 40%] over [period, e.g. 6 months].
Topic: [topic: standard | YMYL (money, health, legal, safety)]
</context>
 
<task>
Audit the article for content decay and give me a refresh
plan. Don't rewrite it.
</task>
 
<requirements>
- Compare intent and format with the current top results:
  has the SERP changed what it rewards?
- Sections to update, add, merge or cut, each with a reason.
- Outdated facts to verify; on a YMYL topic every factual
  claim needs a primary source.
- A new title tag and meta description; internal links to
  add. Keep the URL and the sections that still rank.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Each action ties to a gap against the top results or to
  new information.
- The plan says what the refreshed page will offer that the
  top results lack.
- Nothing that still earns traffic is cut.
</quality_bar>
 
<output_format>
A table: Section | Action (keep, update, add, merge, cut) |
Reason | Evidence. Then facts to verify, the new title tag and
meta description, and links to add.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. End with a list of the
assumptions you made.
</process>

List every checkable claim in this draft (numbers, dates,
names, quotes, features) with the source text that supports
it, or UNSOURCED. Then missing disclosures and AI tells.
Don't fix anything. Check: no claim is skipped.
[paste sources, then the draft]

<materials>
Sources: [paste]
Draft: [paste]
</materials>
 
<context>
Channel: [channel, e.g. blog and LinkedIn]
Connection: [connection: none | paid | free product | affiliate]
Media: [media: none generated | AI images | realistic AI video]
Topic: [topic: standard | YMYL]
</context>
 
<task>
Run the pre-publish check: claims, disclosures and AI tells.
Report only; I'll make the fixes.
</task>
 
<requirements>
- Claims: every number, date, name, quote, product feature
  and cause-and-effect statement, with the exact sentence and
  the supporting source quoted, or UNSOURCED. Then claims
  that are sourced but overstated.
- Disclosures: a material connection stated in the post
  itself (FTC); platform labels for realistic synthetic
  media (YouTube requires one); our own AI-use line.
- AI tells from this sheet's list, quoted, with the tell.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- No sentence with a number, name, date or quote is missing
  from the claims list.
- Each supported claim quotes the source, not a paraphrase.
- Each disclosure finding names the rule behind it.
- Nothing is rewritten.
</quality_bar>
 
<output_format>
Three tables: Claim | Source quote or UNSOURCED | Overstated?;
Disclosure | Needed? | Where it goes; Tell | Quote | Fix
direction.
</output_format>
 
<process>
End with a list of the assumptions you made and anything you
couldn't check from what I pasted. This is not legal review.
</process>

Write an SEO-optimized blog post about time tracking for
freelancers.

<materials>
Survey: [paste our survey of 1,200 freelancers]
Top results: [paste the top 10 titles and URLs for the query]
Voice guide: [paste]
</materials>
 
<context>
We make a time-tracking and invoicing app for solo
freelancers. Readers: freelance designers and developers in
their first three years who undercharge because they don't
track time. The article targets "how to track billable
hours": informational intent, top of the funnel. The top
results are generic how-tos and app roundups.
</context>
 
<task>
Outline a pillar article that wins this query by arguing that
you track time to invoice accurately, not to be more
productive. The outline is for my approval before drafting.
</task>
 
<requirements>
- Intro answers the query in 2-3 sentences.
- H2s: what counts as billable; three tracking methods
  compared in a table; turning tracked time into an invoice;
  common mistakes. H3s for searchers' sub-questions.
- Under each heading: the question it answers, key points and
  the survey result that supports it.
- Mark places for the author's own invoicing records
  [AUTHOR TO FILL].
- Internal links: /pricing and /guides/invoice-template, with
  descriptive anchors. One CTA, at the end: start a free
  trial.
- Banned: "delve", "game-changer", "in today's", rhetorical
  questions.
- Never invent numbers, names, quotes or sources: write
  [NEEDS SOURCE] where support is missing.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The angle (invoicing, not productivity) holds in every H2.
- At least 3 points the top results lack, each from the
  survey or the author's records.
- Every number traces to the survey or carries
  [NEEDS SOURCE].
- A skimmer gets the method from the headings alone.
</quality_bar>
 
<output_format>
The outline as nested bullets. Then a table: 5 title tags
(50-60 characters, with counts), the slug and a meta
description of about 150 characters.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would
change your answer; otherwise go on. Stop after the outline.
End with a list of the assumptions you made.
</process>

H1: How to track billable hours so every one gets invoiced
 
Intro: log time against a client and task as you work, then
reconcile the log with your invoice each week. Proof: survey
Q7 (days to invoice, trackers vs non-trackers).
 
H2 What counts as billable
  - Calls, revisions, research done for the client
  - Gray areas: travel, admin, scope creep
    [AUTHOR TO FILL: your rule for revision rounds]
  - Proof: survey Q3 (tasks freelancers forget to bill)
 
H2 Three ways to track billable hours, compared
  - Table: three popular time-tracking apps by price, free
    tier, invoicing features and integrations
  - Close: why our app suits solo freelancers
 
H2 Turning tracked time into an invoice
  - Round to 15 minutes or bill exact minutes?
    [NEEDS SOURCE: norms in your field]
  - Line items clients don't dispute [AUTHOR TO FILL]
 
H2 Common mistakes
  - Rebuilding the week from memory on Friday
  - Skipping "small" tasks; proof: survey Q5
 
Title tags: How to Track Billable Hours (and Bill Them) (43)
...
Slug: /how-to-track-billable-hours
Meta: What counts as billable, three ways to track it, and
how to turn tracked time into invoices clients pay. (112)
 
Assumptions: readers bill hourly, not per project; the
survey questions are numbered as in the attached file.

Good outline. Change only H2 2: compare three tracking
methods, not apps: a running timer, an end-of-day timesheet
and rebuilding the week from a calendar. Columns: method |
accuracy | effort per day | best for | survey evidence. No
app or brand names in this section. Lengthen the meta
description to about 150 characters with the payoff. Keep
every other heading word for word, then draft the intro and
H2 1.

Compare this draft with the current top results I pasted.
What does it say that none of them say? If nothing, propose
three additions based only on my source material.

Write 15 hooks for [platform: LinkedIn | X | Shorts]: 3 each
of specific number, contrarian, mistake, story in progress
and question, all true to the content.
Rank your top 5, one line each on the tension it creates.
[paste the piece]