73 real seed-stage AI pitch decks with teardowns and measured benchmarks: median 13 slides, 10–15 typical, verified round sizes.
Seed AI Pitch Deck Examples & Benchmarks
Every published AI teardown whose own filed stage note says seed — 73 decks from 72 companies. The benchmarks below are measured from exactly those decks.
Measured on 73 published teardowns (72 companies) — the same decks listed on this page. Slide counts and years come from each deck's own filed facts; round sizes only from rounds we could verify. Decks missing a figure are excluded from that figure's denominator rather than counted as zero.
The measured numbers
Decks in this cohort: 73 — 72 companies, each with a published teardown
Median deck length: 13 slides — Middle half runs 10–15 slides (73 decks with a slide count)
Median verified raise: $3.5M — 55 of 73 decks have a verified round size
Decks from 2022 or later: 32 of 73 — 23 predate 2018; every deck states its year
What the numbers say
Half of these decks run between 10 and 15 slides, with a median of 13. The shortest is 7 slides and the longest 44.
That is 1 slide shorter than the wider AI & Machine Learning collection, whose median is 14 across 200 decks.
That median matches the wider Seed & Pre-Seed collection (13 slides across 1,265 decks).
37 of the 55 decks with a verified round raised $1M–$10M, the most common band here. 18 decks carry no verified figure and are left out of every raise number on this page.
The cohort leans recent: 32 decks are from 2022 or later against 23 from before 2018.
Where slides tend to land
Slide labels exist for 70 decks in this cohort, and they cover only part of each deck. So the table below reports where a slide type was found, not how many decks contain one — a slide we never labelled is not a slide the founder omitted.
Solution — found in 28 of 70 labelled decks, typically slide 5
Team — found in 23 of 70 labelled decks, typically slide 3
Problem — found in 22 of 70 labelled decks, typically slide 4
Product — found in 20 of 70 labelled decks, typically slide 5
Median length is 13 slides, with the middle half between 10 and 15 — shorter and much tighter than the 11–19 band across all AI decks.
At seed there is less to evidence, and the decks that got funded did not pad the gap.
What to do with it: If your seed AI deck is over 16 slides, the extra slides are almost certainly answering questions nobody has asked yet.
55 of the 73 decks have a verifiable round size, and the median is $3.5M.
That is the size of round these decks were actually used to raise — a useful sanity check against an ask that is three times larger with the same evidence behind it.
What to do with it: Size your ask against the milestone it funds, then check it against this median before you send.
Among the 70 labelled decks, solution appears most often (28) around slide 5, and team appears in 23 around slide 3 — earlier than any other cohort we measure.
At seed the team is a substantial part of the argument, and these decks front-load it.
What to do with it: Move team forward. If your team slide is at the back, a seed reader reaches your traction slide without knowing why you are the right people to trust with it.
Limitation: Labelling is partial; these are counts of where labelled slides were found, not coverage percentages.
What seed investors publish about AI pitches
Seed guidance from the accelerators and funds themselves, linked to source. Use it for expectations; use the numbers above for what decks do.
Raise to a milestone, not to a feeling. Y Combinator's seed guide frames the amount as a function of what it buys, and gives a concrete cost anchor — roughly $15k per engineer per month, all-in — so founders can defend the number. (Y Combinator 2023)
Run the round with a start and an end date. Techstars advises founders to set clear start and end dates so the round keeps momentum rather than dying on the vine — a discipline that also shapes how much traction the deck needs to carry. (Techstars 2025)
Say the number, and what it gets you. Y Combinator's guidance on talking to investors is to make the numbers clear and meaningful, and to tell the investor how much you need and what it buys. (Y Combinator 2018)
Rounds are taking longer between raises. Carta measured the median wait between rounds at 696 days in Q2 2025. A seed deck that assumes an 18-month runway to the next raise is planning against a shorter cycle than the market is currently running. (Carta 2025)
A seed AI deck in thirteen slides
Thirteen is the median in this cohort. This ordering follows where labelled slides land here.
One-sentence definition. What it is, in words a non-specialist repeats correctly. (Sequoia Capital)
Problem. Labelled decks put problem around slide 4; name the person and the current workaround.
Team. Team lands around slide 3 here — unusually early, and it works at seed.
Solution. The most common labelled slide in this cohort, around slide 5.
Product, shown. Real output on screen. At seed this is often the strongest evidence you have.
Why now. The capability or cost change that opened the window. (Sequoia Capital)
Early evidence. Design partners, waitlist, usage or letters of intent — labelled honestly as early.
Ask and milestone. Median round in this cohort is $3.5M; state what yours funds. (Y Combinator 2023)
Material that usually belongs in an appendix rather than the main deck: evaluation results and model detail; pipeline by named account; 18-month hiring plan.
Seed AI decks worth reading
Covers and commentary only; each links to the full teardown.
Pruna AI. Technical company that still keeps the first three slides readable by a generalist. Read the Pruna AI deck teardown
What the stronger seed decks do
Team early. Team around slide 3 lets everything after it be read as evidence about people the reader already trusts.
Small, real evidence. Three named design partners with usage beats a projection chart at this stage.
A defensible ask. The ask is tied to a hiring plan and a milestone, not to a valuation target.
Seed-specific mistakes in this cohort
Presenting a seed idea with a Series B deck structure. It invites metrics comparisons you cannot win and buries the team argument. Instead: Use the 10–15 slide shape this cohort actually uses. Basis: Measured in this cohort (n=73).
An ask with no milestone attached. Y Combinator's guidance is explicit that the number needs to come with what it buys. Instead: State the amount, the runway it buys and the milestone it reaches. (Y Combinator 2018, Y Combinator 2023) Basis: Investor guidance.
Planning the next round 12 months out. Carta measured a 696-day median gap between rounds in Q2 2025, so a 12-month plan assumes a market speed that is not currently there. Instead: Show runway to the milestone with margin, and say what happens if the next round is late. (Carta 2025) Basis: Cited market data.
Seed AI deck checklist
10–15 slides.
Team by slide four.
Product output visible, not described.
Evidence labelled as early where it is early.
Ask tied to a milestone and a runway.
Round size sanity-checked against the $3.5M median here.
Check your seed AI deck against these 73 decks
Pitch Score compares your deck's length, ordering and evidence with the seed AI cohort measured on this page.
The median here is 13, with the middle half between 10 and 15 slides.
How much do seed AI companies raise?
Of the 73 decks in this cohort, 55 have a verifiable round size and the median is $3.5M.
How do I size the ask?
Y Combinator's seed guide ties the number to what it buys and gives roughly $15k per engineer per month as an all-in cost anchor to build the plan from. (Y Combinator 2023)
Do I need traction to raise a seed round in AI?
The decks here vary widely, and many carry early evidence rather than revenue. Techstars frames it as story getting you in the door and evidence proving capability — at seed the balance sits further towards story than at Series A. (Techstars 2025)
How long should I plan for the round to take?
Set explicit start and end dates for the raise; Techstars recommends this to keep momentum. Separately, Carta measured a 696-day median gap between rounds in Q2 2025, which matters for the runway you ask for. (Techstars 2025, Carta 2025)
Methodology and limitations
The cohort is the 73 published deck teardowns on this page (published seed-stage AI deck teardowns). Nothing outside that set is counted, and the set is frozen so the numbers on this page do not move under you.
Slide counts come from the original deck file, not from a summary or a re-typed copy.
Slide-type positions are only computed over the 70 decks in this cohort whose slides are labelled. That is why every row shows "found in X of Y decks" rather than a percentage — a percentage would imply the other decks lack the slide, when in fact we simply have not labelled them.
Round sizes are the amounts publicly reported for the round the deck was used for; decks with no verifiable amount are excluded from that statistic rather than estimated.
External claims on this page link to the publisher that made them, with the date we checked the page. We do not restate a third-party number without a link to its own source.
We publish deck covers and link to each teardown. We do not republish slides from these decks.
