A seed deck with €420K in ARR and 32 paying customers scored 39 out of 100, and the Pitch Deck Scorer listed the eight fixes worth 21 more points.

The Pitch Deck Scorer is a skill you add to Claude: give it your deck and it returns a score out of 100 and a full scorecard as a web page and PDF.

It matters because investors read fast. Papermark's 2026 report on 24,541 decks puts the average view at four minutes, and 16% of views end in under ten seconds.

The Pitch Deck Scorer doesn't stop at the number: it ranks the changes that raise the score most and writes the new headlines for you.

Below is everything the Pitch Deck Scorer does, a full test run and the setup. Enduring Founder Club members can download the skill at the end of this article.

Most deck feedback is a list of opinions. The Pitch Deck Scorer gives you a work plan.

Attach a PDF, a PowerPoint (PPTX, PPT or ODP) or slide images. The skill turns the deck into slide images, text, speaker notes, headlines, every figure with the slide it appears on, and objective metrics such as words per slide, font families and text too small to read. Flattened exports, like Canva PDFs, are read from the images when needed.

You get back:

→ The skill never invents your numbers. Where data is missing, the example uses an [x] placeholder for you to fill in.

It also works in your language. Ask in Spanish and the scorecard comes back in Spanish.

Every variable is rated 0–5 against written anchors, then weighted into three pillars:

The scoring is deliberately strict. When the skill is torn between two levels, it takes the lower one, and it scores only what is on the slides, because an investor reading alone won't know what you know. It is also stage-fair: a pre-seed deck isn't penalized for missing cohort data, and a deeptech seed deck isn't penalized for having no revenue.

Four red flags cap the total, however polished the rest is:

→ A better template won't rescue a deck that loses most of its points on evidence. The pillar breakdown tells you which problem you actually have.

A generic checklist reads every deck the same way. Investors don't.

When the deck is an AI company, the skill switches on an AI lens with the eight questions AI investors ask now: what's built beyond the API, why the labs or incumbents won't ship it as a feature, whether usage improves the product, what inference does to gross margin, how solid the revenue is, whether performance is measured, why now, and what happens if a model provider or a regulator changes the rules.

For deeptech, hardware and bio, a second lens checks technical proof and readiness level, defensible science, the regulatory path, unit economics at scale, path to revenue, capital plan by milestone, execution team and scale-up risk.

The lenses add no bonus points. They sharpen the content scores, and they can trigger a cap: an AI deck that fails both 'beyond the API' and 'lab-proof' cannot score above 59.

In the test deck, the AI lens returned five fails out of eight. The deck named GPT-5 and Claude as its engine but never said what was proprietary, what each review cost to serve, or how it would handle confidentiality and GDPR for customers' contracts.

→ If you're raising for an AI company, these are the questions in the room. Better to meet them in the scorecard first.

To test the skill, we built Clauseboard: a fictional 12-slide seed deck for an AI contract-review startup, loaded with the most common first-draft mistakes. It scored 39/100. Three findings show how the skill reads a deck.

€420K ARR from 32 customers is a strong seed signal. The deck put it sixth, titled "Traction", above a chart with no dates, values or axis labels.

The skill ranked rebuilding this slide as the top fix, worth an estimated +6.2 points: monthly ARR with labelled months, the growth rate, paying customers separated from pilots, and the slide moved straight after the problem. Reading data backs the move: Papermark found that decks opening with traction in the first three slides get 20% more views.

Slide 6 said €420K ARR. The vision slide said "over €500K". The skill caught the mismatch by comparing every figure across slides, and capped the deck at 59 until it is fixed with one figure, one definition and one date.

The market slide sized Clauseboard against the $1.8 trillion global AI market. The skill scored it 1 out of 5 and wrote the replacement structure: target companies × seats per company × €99 per month × 12, with sources.

→ The single biggest fix was worth more than the next two combined. That ranking is the point of the report: it tells you where an hour of work moves the score most.

A fair question: if the bar is this high, how do weak decks get funded?

The skill answers it in its own calibration. To keep scores consistent, it compares every deck against public decks scored on the same anchors. Airbnb's 2009 seed deck lands at an estimated 55–65: a one-sentence explanation of the business by slide 2 and named competitors, but its own traction barely shown. Scalestack's 2023 deck, behind a $1M AI sales-tech seed round, lands at 50–60: strong narrative, no pricing model or go-to-market. Terra One's 2024 deck, behind a $7.5M seed round in energy hardware, lands at 35–45: a powerful problem slide, but no competition slide and no ask.

All three raised money. None would score above about 65. That's expected: strong founders, warm intros and hot markets carry weak decks.

→ The score measures how well the deck makes the case on its own, not whether your company is fundable. Its job is to make sure the deck is never the reason you get a no.

The Pitch Deck Scorer runs in Claude, on Free, Pro, Max, Team and Enterprise plans.

If you don't say how the deck will be used, the skill assumes it will be read alone, the harder test, and tells you what it assumed.

Your deck stays private: the skill is instructed never to put your numbers, customers or company name into web searches. And if code execution is off, it still scores the deck and gives you the fixes in the chat; you only lose the formatted scorecard.

The Pitch Deck Scorer is part of the Enduring Founder Club: practical fundraising tools for founders building for the long term, limited to 100 members.