The argument in one line.
The most valuable use of AI in email marketing is not writing the email but auditing it against a doctrine of human-language rules you upload to a Claude Project -- turning a general model into a brutally specific copy critic.
Read if. Skip if.
- You send cold or warm emails for coaches, consultants, or course creators and are getting low open or click rates.
- You have used AI to write emails and noticed the output sounds generic or robotic.
- You are a freelance email copywriter who wants a systematic pre-send checklist you do not have to run manually.
- You want to understand which specific word patterns make email sound AI-generated at the sentence level.
- You already have a rigorous human-editor review process and are not interested in AI-assisted proofing.
- You are looking for an email builder or automation platform, not a copy-quality audit tool.
The full version, fast.
The creator built a Claude Projects tool called HLOS humanizer score by uploading 14 sections of copywriting doctrine -- lists of forbidden buzzwords, robotic transitions, weak CTA patterns -- then wiring in the Anthropic API. When you paste an email and your goal, it returns a 0-100 score, counts of buzzwords and filler words, a readability grade, structural issues with line-level fixes, a performance critique across clarity, specificity, persuasion, and CTA strength, plus a rewrite template. A demo email written entirely by AI scored 17/100. The practical takeaway: Claude Projects can become a domain-specific copy auditor when you supply it with explicit scoring criteria, rather than letting it generate copy that sounds like every other AI-written email.
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.

01 · Credentials opener and promise
Establishes Upwork credibility (70k+, 100+ clients), states what the system can do: score emails, flag AI buzzwords, filler words, readability, CTA strength, and produce a rewrite template.

02 · What the system is NOT for
Explicitly warns against using the tool to write the email; positions it as an auditing and strengthening tool only.

03 · Getting an Anthropic API key
Screen-walks console.anthropic.com, creating an API key, and mentions the free plan.

04 · Loading the tool and pasting a demo email
Pastes API key into HLOS UI, loads a demo AI-written email with subject line and goal (book a discovery call), hits analyze.

05 · Inside the skill docs (HLOS system)
While analysis runs, shows the Google Docs uploaded to Claude Projects: 14 sections of AI word patterns, robotic transitions, filler phrases, weak subject lines and CTAs.

06 · Google Drive folder reveal
Shows the full stack email marketing skills folder offered as a lead magnet for newsletter subscribers.

07 · Reading the live analysis (17 out of 100)
Walks through the live output: 19 buzzwords, 7 filler words, readability 45/100, 116-word count, flagged word list, structural issues, performance critique, brutally honest verdict.

08 · Rewrite template output
Shows the fill-in-the-blank rewrite template the system produces, with subject line tied to a specific pain point and annotated placeholders.

09 · Second tool and CTA
Mentions a buzzword-only scanner (no performance critique or rewrite), directs viewers to newsletter for both tools.
Lines worth screenshotting.
- An email that scores 17/100 on an AI audit almost always fails on the same axis: every sentence could belong to any product sent to any person.
- The fastest single edit to raise email readability is removing filler words -- really, very, quite, actually, basically -- which lower perceived authority without adding meaning.
- Words like unlock, revolutionize, streamline, and optimize have been so overused in AI copy that readers have trained themselves to stop reading when they appear.
- Claude Projects becomes a domain-specific copy auditor when you upload explicit doctrine -- forbidden phrases, structural rules, CTA patterns -- rather than relying on its general knowledge.
- A subject line for a warm list must earn its open with specificity tied to the reader's current situation; curiosity hooks designed for cold traffic fail on subscribers who already know the sender.
- Click here to learn more as a CTA is systematically mismatched to any high-ticket offer -- the micro-commitment action conflicts with the high-trust outcome it is meant to drive.
- Feeding AI a pre-built evaluation rubric produces feedback measurably more specific than asking the same model to improve this email with no criteria.
- The rewrite template the tool produces is not a finished email but a fill-in-the-blank scaffold showing exactly where specificity needs to be inserted and what kind.
- Building a copy-audit tool as a Claude Project costs nothing beyond an Anthropic API key and the time to document your own copywriting rules.
- Comment-gating and newsletter-gating a useful tool turns a tutorial video into an active lead-generation asset.
How to turn Claude into a copy editor that scores every email.
The leverage is not prompting Claude to write emails but uploading your own doctrine -- forbidden words, structural rules, CTA patterns -- so the model has explicit criteria to score against instead of producing generic feedback.
- An email that scores 17/100 almost always fails on the same axis: every sentence could belong to any product sent to any person -- specificity is the only fix.
- The fastest single edit to raise readability is removing filler words (really, very, quite, actually, basically, literally) which lower perceived authority without adding meaning.
- Words like unlock, revolutionize, streamline, and optimize have been so overused in AI copy that readers have conditioned themselves to stop reading when they appear.
- Feeding Claude Projects a comprehensive doctrine document turns a general-purpose model into a domain-specific copy auditor -- quality of critique scales with specificity of the uploaded docs.
- Subject lines for warm lists need a reason tied to the reader's current situation; curiosity hooks designed for cold traffic fail on subscribers who already know the sender.
- Click here to learn more as a CTA is systematically mismatched to high-ticket offers -- the micro-commitment action conflicts with the high-trust outcome (discovery call) it is meant to drive.
Things they pointed at.
Lines you could clip.
“Writing the email will just destroy the whole overall strategy of this AI system and I do not want you to concentrate on writing the email.”
“This system is actually trained for it to be brutally brutally brutal with you, with your email.”
“This email will not work. It reads like a mail merge that was never finished.”
Word for word.
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
The bait, then the rug-pull.
A freelance email marketer with over 70k in Upwork earnings opens not by showing results but by handing you the system -- a Claude Projects audit tool that scores your email 0-to-100 and flags every buzzword before you hit send.
Named ideas worth stealing.
HLOS (Human Language Operating System)
A Claude Projects configuration consisting of 14+ sections of uploaded doctrine: forbidden AI buzzwords, robotic transition patterns, filler word lists, weak CTA patterns, subject line rules. The model reads these files and scores any pasted email against them.
Email Performance Critique dimensions
- Clarity
- Specificity
- Persuasion strength
- CTA strength
The four axes the HLOS system scores each email on, each rated and explained with specific line-level feedback.
How they asked for the click.
“Just click in the description right there. You will find the newsletter. And in the newsletter, when you just sign up, you will get it as a welcome email.”
Double-gated: primary CTA is newsletter signup for the skill docs folder; secondary is comment-gating (ask in comments to get the HLOS system link directly). The lead magnet is the tool itself.







































































