Here is what you will need to run these skills, automations, or AI workflows:
1. A Claude Pro account
2. The Claude in Chrome extension
3. The Claude desktop application (required to run Cowork)
4. An n8n account for automation
5. A Groq account API for the n8n automation connection
6. Please log in to your Instagram account
The first step you will need to take is to download Claude on your desktop, start the application, and connect it with Notion. Once that is set up, you can begin running the skills. If you are connecting an ad account, you can do that now; otherwise, proceed directly to the skills.
Available skills include:
(a) Review mining
(b) YouTube analysis
(c) Reddit skill
(d) Research or web query skill
(e) Instagram trends
(f) Content gap analysis
(g) Full package or pipeline skills
You will also need to provide a Notion database where all the data and results will be stored.
For these tasks, I recommend using the Opus 4.6 or 4.8 models. Please be aware that these models consume a significant amount of credits, which is important to keep in mind. While you could try using the Sonnet model, it does not work as deeply and simply won’t provide the same quality of output as Opus 4.8 or 4.6.
As a final step for creating viral scripts for ads, I recommend feeding the AI data regarding your specific voice and tone. This ensures the AI works in your personal style to produce the most effective scripts.
This is the comprehensive workflow for creating high-performing ad content using deep research and automation.
1. Ad Account Connection
The first step is connecting your Facebook ad account to Claude. If you haven't done this yet, go to the connections settings. A quick Google search can guide you through the process if needed.
(a) If you prefer not to connect your account, the workflow is still relevant.
(b) However, you will need to manually provide your current winning ad angles to Claude. Without this data, Claude won't understand what is currently performing well for your brand.
2. Review Mining
We will use a specific skill (running on Claude) to mine review sites like Trustpilot, G2, and Amazon.
(a) The system will analyze what customers are actually saying in their reviews.
(b) Based on this sentiment, it will generate ad angles and save them to a Notion database or a Markdown file.
(c) You simply provide the product details, and the tool performs the deep research for you.
3. YouTube Analysis
The next stage involves exploring YouTube to identify trending topics and niche-relevant discussions.
(a) The tool checks both fresh videos and older, "evergreen" popular content.
(b) It identifies comments with the highest engagement to understand the audience's mindset.
(c) These insights are then compiled into your Notion database or Markdown file.
4. Reddit Research
The system will conduct comprehensive research on Reddit to analyze your topic, product, or target audience.
(a) It finds the most relevant subreddits and threads with the highest upvotes.
(b) It extracts high-engagement comments to give you a clear picture of community sentiment.
5. Web Query & Community Forums
We will run a broader web query to find what people are talking about in various online forums and communities. To avoid redundancy, this step specifically excludes the review sites and Reddit threads already covered in previous steps.
6. Instagram Trends
This skill explores viral trends and reels on Instagram. It identifies what is currently going viral to help you stay ahead of social trends.
7. Content Gap Analysis
This critical step identifies what you are currently missing. It compares all the data collected from the previous steps with your current ad account performance.
(a) If your ad account is not connected, you must provide your current angles manually for this to work.
(b) The system performs in-depth research on the collected database to find "uncovered" angles—things your competitors aren't doing or topics not being covered in depth.
Final Result:
Once these steps are complete, you will have a comprehensive list of researched angles. We then take the scripts from viral videos and recreate new scripts using these proven viral styles, tailored specifically to your product. This ensures your final content is based on strong, comprehensive research and proven performance styles.
n8n
Step 1 for n8n - create account
Step 2 Click on create workflow
Step 3 - Paste the json code here
You can get the code here.
This will appear after pasting.
Step 4 - Create an account on Groq and create an API key and copy the API key
Step 5 - Paste the API key in n8n
Step 6 - Next copy the n8n web hook (the production URL)
Research Skills — User Guide
Web Ad Researcher
What it does: Scans the open web to collect raw, unfiltered audience language — forums, Reddit-style communities, Quora, news articles, and blogs. The goal is to find what real people say about a product or problem when they are not being marketed to.
Why it matters: Complaint threads, late-night forum posts, and Quora questions contain more emotionally charged language than any editorial source. That language is the raw material for ad hooks and copy angles.
How it works:
Asks for the product, niche, and target audience before doing anything
Builds 20 to 30 search queries across six source angles: forums, Quora, niche communities, news, blog reviews, and deliberate complaint queries
Evaluates each result for recency, engagement, and authenticity
Skips thin or promotional content automatically
What the output includes:
Per source:
URL, source type, platform, date, engagement signals
Tagged extractions: pain points, desires, objections, proof statements, questions, hook candidates, and angle suggestions
Master output:
Ad Angles Master List with 8 to 15 distinct creative directions
Each angle includes a hook draft, supporting evidence, and a suggested creative format
What to prepare before running:
A Notion database URL (or request a Markdown file instead)
A clear, specific description of who the audience is
YouTube Ad Researcher
What it does: Mines YouTube for two things: how creators frame content to get views in your niche, and what real people say in the comment sections beneath that content.
Why it matters: YouTube comment sections are one of the highest-signal sources available. People who comment on a 10-minute video about a specific problem are invested, opinionated, and unguarded. Comments with 10 or more likes show shared feelings, not isolated opinions.
How it works:
Builds a keyword set across five angles: pain points, product and solution terms, competitor names, creator/educator keywords, and social proof or transformation-style search terms
Filters videos by view count and recency
Reads the top comments on each qualifying video
Tags each comment: pain point, desire, objection, proof statement, question, hook candidate, or angle signal
Targets 10 to 30 comments per video, across 10 to 40 videos depending on the depth selected
What the output includes:
Per video:
Summary, key observations, and extracted comments sorted by category
Master output:
Full Ad Angles report
Ranked pain point list and ranked desire list
Objection bank
Raw hook candidates bank
Table of high-value channels to monitor
What to prepare before running:
A Notion database URL
A precise audience description including age, specific problem, and awareness level
Reddit Researcher
What it does: Goes deep into Reddit's community structure to collect the most candid audience language available in any research skill.
Why it matters: Reddit communities form around shared identities and shared problems, not around brands. People post on Reddit when they are genuinely stuck or frustrated, and they do it anonymously, which means they say things they would never say in a survey, a review, or a public social post.
How it works:
Builds a keyword map across four dimensions: problem keywords, solution keywords, identity keywords (how the audience describes themselves), and outcome keywords (what they want to achieve)
Uses those keywords to find the 5 to 10 most relevant subreddits
Pulls threads that match specific signals: requests for help, expressions of frustration, transformation stories, and posts questioning whether something actually works
Reads comments and extracts pain points, desired outcomes, objections, false beliefs, trigger events, emotionally charged phrasing, and competitor mentions
Prioritises comments with 50 or more upvotes
What the output includes — four Notion tabs:
Overview page: Research scope and top insights
Raw research tab: One entry per Reddit thread with verbatim quotes and context
Synthesised intelligence tab: Pain point map, desire map, objection bank, trigger event library, verbatim language bank, and worldview audit
Ad angles tab: 8 to 12 fully written angle blocks, each with a strategic frame, target emotion, three hook variations, body direction, CTA guidance, and Reddit evidence
What to prepare before running:
A Notion database URL
A specific and narrow audience brief (this skill produces the most detailed angle output, so precision matters most here)
Review Researcher
What it does: Mines verified buyer reviews across Trustpilot, Amazon, G2, Capterra, the App Store, Google Play, and Yelp to extract real customer language for ad creative.
Why it matters: Review writers actually bought and used the product. They report on a lived experience, often with strong emotion attached. 1-star and 5-star reviews are equally valuable: a 5-star review describing life after purchase is transformation copy; a 1-star review about what went wrong is objection copy; a 3-star review is the most nuanced signal of all because it shows exactly what a buyer wanted but did not get.
How it works:
Asks for: product, niche, audience, which platforms to research, competitor brand names to include, output preference, and research depth
At Deep depth, targets 60 to 100 or more individual reviews across all platforms
Segments reviews by star rating before extracting anything
Tags every extraction: pain, desire, emotion, outcome, objection, competitor comparison, feature request, hook candidate, or red flag
Keeps the star rating attached to every quote
Prioritises reviews other buyers found most useful (by helpfulness vote count)
What the output includes:
Per review source:
Verbatim reviews with tags, patterns within the batch, and a note on which ad angles the batch supports
Master output:
Full lists of pain points, desires, emotions, objections, competitor mentions, hook candidates, transformation stories, and unexpected benefits
Every entry includes platform, star rating, date, and a direct URL back to the original review
What to prepare before running:
A list of competitor brand names if you have them — comparison signals between your product and a competitor are among the most actionable findings this skill produces
What it does: Goes to TikTok and Instagram to find what is generating real organic reach right now in your niche, then extracts the hooks, formats, and audience signals behind that virality.
Why it matters: A hook that earns two million views without any media spend has been market-tested at scale in real time. That is more current proof of what a paid audience responds to than any focus group or historical research source.
How it works:
Accepts optional specific creator profiles on either platform so you can research accounts already performing well in your space
Builds a keyword and hashtag set across four dimensions: problem and pain keywords, solution and category keywords, identity and audience keywords, and trend-style phrasing
Works through TikTok and Instagram separately
For profile-specific research: calculates each creator's baseline view average (excluding top outliers), then flags anything at 5x that baseline as a viral outlier
For every qualifying piece of content, captures: direct URL, view count, likes, comments, the exact hook from the opening seconds, a format tag, and angle tags
Mines top comments on the highest-performing content for verbatim audience language
What the output includes:
Per piece of viral content:
Source link, engagement figures, extracted hook, format classification, angle classification, top comments with engagement counts, and a note on which ad angles it supports
Master output:
Ad Angles Master List with hook drafts and evidence
Hook Bank of verbatim hooks with source URLs
Format Intelligence table showing which creative formats dominate the niche right now
Trending Audio section (flagged with time-sensitivity warning)
Audience Language Bank of comment phrases for ad copy
What to prepare before running:
A specific topic description (not "health supplements" but "collagen supplements for women over 40")
Any creator handles you want researched
Content Gap Analysis
What it does: Reads the databases the other five skills have built and identifies the gap between what your audience is saying and what your current ads are actually addressing.
Why it matters: Individual research skills collect signals. This skill compares all of them against your existing creative and tells you exactly what you are missing, ranked by priority and backed by source evidence.
How it works — three modes:
Mode 1 (connected ad account or Notion ad database):
Loads your live ads automatically
Reads hooks and primary text
Builds a Current Coverage Map showing which angles, pain points, desires, objections, emotional registers, and formats your ads already cover
Loads all available research databases, builds a Research Universe, and runs a direct comparison to find everything your ads are missing
Mode 2 (no ad account connected):
Same setup as Mode 1
Instead of pulling ads automatically, asks you to describe your current angles or paste existing ad copy directly into the conversation
Builds the Coverage Map from what you provide
Mode 3 (no ad account or angle input needed):
Runs a deep dive into the Review Research and Social Trend databases
Produces a complete opportunity map covering: high-signal pain points not in any ads, emotional registers not being used, desires not yet promised, objections not pre-empted, competitor gaps not exploited, unexpected benefits not featured, verbatim hook candidates, transformation stories ready for UGC briefs, viral angles not in current ads, dominant formats the paid mix is missing, trending audio opportunities, and platform-proven hooks from real high-view organic content
What the output includes:
Executive summary
Priority 1 gaps (immediate action)
Priority 2 gaps (test soon)
Priority 3 gaps (monitor)
Coverage confirmation section for angles already research-backed
A full creative brief for every Priority 1 and Priority 2 gap
Each gap entry includes: gap type, signal tier, verbatim quotes with platform and URL, what current ads say (or don't say), why the gap matters, a suggested angle with a hook draft, body direction, and format recommendation.
Confidence weighting:
A pain point appearing in reviews, Reddit, YouTube, and a viral TikTok simultaneously gets maximum-confidence status
Social trend data carries extra recency weight (it shows current platform behaviour, not historical sentiment)
Trending audio gaps are flagged with a time-sensitivity notice
What to prepare before running:
At least one research database that is built and up to date
If using Mode 2: your current ad angles or some live hooks written down
Full Ad Research Pipeline
What it does: Runs your entire market research process from start to finish, across every major source where your audience speaks candidly. One skill, all five research sources, one connected output.
Sources it covers:
Verified buyer reviews on Trustpilot, Amazon, G2, and the App Store
Reddit threads
YouTube comment sections
Forums, Quora, news articles, and blogs across the open web
Viral Instagram Reels (hooks, formats, and angles getting traction right now)
What gets built per source:
A structured Notion database
Real audience quotes
Tagged pain points, objections, and desires
A prioritised list of ad angles with hook drafts and creative format recommendations
Direct links back to every original source
What happens after the research is built:
Once all five databases are complete, the skill runs a Content Gap Analysis automatically. It compares everything found against what you are currently running in your ads.
The gap analysis produces:
Every angle your audience is expressing that your creative is not addressing
Each gap ranked by how many sources confirmed it
A ready-to-use creative brief for every priority gap
What each creative brief includes:
Hook draft
Body direction
Suggested format
Platform recommendation
A callout for any trending Instagram audio worth acting on quickly
Final pipeline output:
One master report summarising every gap
The strongest verbatim hook candidates from across all five sources
A clear set of next steps for your creative team
Ad Scripter
What it does: Takes research you have already gathered and turns it into a finished, platform-ready ad script. No guessing. No writing from scratch.
What it pulls from:
Reddit threads
Trustpilot reviews
YouTube comment patterns
Web query themes
Instagram high-view content analysis
How it works:
You paste in your raw research (even a rough mix of review quotes and Instagram notes is enough to start)
The skill extracts the strongest angles from the material: exact pain points people describe, outcomes they want, things they have already tried that failed, and objections stopping them from buying
It ranks those angles
It selects the right ad framework for each one
Ad frameworks it chooses between:
Problem-Agitate-Solution (best for complaint-heavy research)
Before-After-Bridge (best when reviews are full of transformation stories)
Plus others depending on what the research shows
What it will ask you before writing:
What you are advertising
Which platform the ad is going on (Meta, TikTok, YouTube, or LinkedIn)
Which format you want: UGC, direct response, testimonial, demo, or educational
What the output includes:
Full script broken into labelled, timed sections
A note on which research source each angle came from
The specific objection the script is designed to pre-empt
What you can request after the first script is approved:
Hook variations
A shorter cut
A second script built around a different angle from the same research batch
Before Running Any Skill
All eight skills follow the same opening pattern. They ask a small number of setup questions first, then store your answers for the rest of the session.
Minimum inputs required across all skills:
Product or topic description
Target audience description
Notion database URL for the output
Output format options:
Notion database (preferred — lets all research databases connect to each other and feed the Content Gap Analysis)
Markdown file (fallback if no Notion database is set up)
Research depth:
All skills default to Deep if you do not specify
Deep produces more sources, more extractions, and more angles, but takes longer
Quick depth is for fast topic orientation, not comprehensive research
Social Trend Researcher