Clarity Leaderboard

Scroll

Layer 1 · Scored October 2026

How clearly scroll.io communicates, read from the live homepage and re-scored monthly. This measures the website, not the protocol.

Clarity Score

56/ 100

Getting There+2 this month

Score over time

69 → 565 readings since Jun 2026

Rank

#77

of 80 projects

Position

Lost in Translation

Clarity 62 · Narrative 51Below 75 on both axes

Where it sits

Scroll against the other 99, placed by signal clarity across and narrative completeness up.

Ethereum Highloop Akash Network Avalanche Aave Filecoin Helius Immutable io.net Sui Drift Protocol Berachain Celestia EigenLayer GMX Linea Polygon Ronin Arbitrum Fetch.ai Pyth Network zkSync Algorand Core DAO Hyperliquid Internet Computer Monad Optimism Polkadot Stellar VeniceAI Cosmos Elrond (MultiversX)Render Network Sei Uniswap XRP Ledger Celo Ondo Finance Stacks Base Cardano Chainlink Flow Hedera Near Protocol Sky (MakerDAO)Tokeniser Ethena Euler Finance Gnosis Chain Kaspa Lido Manta Network Phantom Solana BNB Chain Flare Helium Injective Jito Pump Fun Secret Network Kite AI Mantle StarkNet Tezos Wormhole Arweave Kamino Finance Tron Aptos Blast Jupiter Pendle Polymarket Scroll Metis Toncoin Bonk Electric Capital 1kx Galaxy Digital Pantera Capital Union Square Ventures a16z Crypto Delphi Ventures Framework Ventures Coinbase Ventures Multicoin Capital Blockchain Capital Haun Ventures Jump Crypto Variant Fund Digital Currency Group Dragonfly Hack VC Polychain Capital Standard Crypto Paradigm

Is this Scroll?

Claim the profile to unlock your private breakdown, see the gap to the companies above you, and run Quick Wins.

A clean ZK gateway story undercut by a broken headline.

Clara · October 1, 2026

Here's what I see: Scroll has a genuinely interesting product idea, one unified interface to route across 30+ AI providers, with zero-knowledge proofs so your prompts stay private. That's a real position. The problem is the page barely lets me feel it.

The very first thing I read is "Your Gateway to Fronti<% M$~&.<" which is either a rendering error or a motion frame caught mid-glitch. Either way, the single most important line on your site is garbled, and that is a trust problem on a page whose entire pitch is trust and verification.

Underneath the diagram, the privacy promise is strong: "no prompts stored · no keys exposed · provider blind." That line does more work than almost anything else here. But most of the page is labels and model names floating in a visual system, not sentences that explain why I should care.

Clarity Score

56/100

Getting There

A clean ZK gateway story undercut by a broken headline.

Messaging Shape

My honest take

The strongest thing you've got is the privacy triad: "no prompts stored, no keys exposed, provider blind." That is specific, it is differentiated, and it is the one place where I understand what Scroll actually protects me from. I'd build the whole page around that if it were me reading this honestly.

But the hero headline is broken. "Your Gateway to Fronti<% M$~&.<" reads as a bug. On the clean full-page render it resolves to "Your Gateway to Frontier Models," which tells me this is a motion or glitch effect. Here's my concern: on a page selling verification and proof, a headline that looks corrupted is the worst possible first impression. It makes me question whether the product works.

Most of the page communicates through a diagram, not through language. ENCRYPT, PROVE, VERIFY, ROUTE, ZK PROOF, a list of model names. These are labels. They gesture at a process but they never explain it in a sentence a developer could repeat to a teammate. I see the machinery but not the meaning.

The product trio is a nice beat: "An app, an API and a box on your desk. Same keys, same models, same proofs." That's distinctive and even a little witty. But it arrives late and gets buried under interactive components (Compass, the sphere, the chain animation) that are described by their loading states rather than their value.

What I don't see anywhere: who this is for named explicitly, why I'd choose Scroll over calling providers directly or using an existing router, any proof beyond the claims, and what happens if I keep routing AI calls the leaky way I do now. The architecture is visible. The argument is not.

Why this keeps happening

I think the page is built visual-system-first, the diagram and the interactive toys came before the words, so the story is carried by labels instead of sentences. That's a sign there's no codified brand strategy underneath deciding what the single most important claim is, who it's for, and in what order to earn their trust. When the strategy isn't set, you default to showing the machinery and hoping it speaks for itself.

What your website actually says

What does your site communicate?

Scroll is a single interface to route across 30+ AI model providers, with zero-knowledge proofs keeping your prompts and keys private.

Who is it for?

Developers and technical teams building on multiple LLMs who care about privacy and provider independence, though the audience is never named.

What problem does it solve?

Juggling many AI providers and exposing your prompts and keys to them, solved by one unified, provider-blind interface.

What outcome does it promise?

One key, every model, with verified private calls and nothing stored or exposed.

Visual signal

contradictorynear-white minimalist palette with a refined serif headline, connoting calm, premium, technical restraint

I like the calm, premium register you've chosen, but the garbled hero text sitting inside it fights the whole "verified and trustworthy" message, so the first thing I feel is a bug, not polish.

Detailed Analysis

One sharp line surrounded by labels and a broken headline.

The copy has a few genuinely good lines. "no prompts stored · no keys exposed · provider blind" is tight, specific, and distinctive. "An app, an API and a box on your desk. Same keys, same models, same proofs." has real voice and rhythm. These moments show someone who can write.

But the primary headline renders as "Your Gateway to Fronti<% M$~&.<", which is either a glitch effect or a bug. In the clean render it reads "Your Gateway to Frontier Models," which is a serviceable but interchangeable infrastructure headline. Even at its best, it names a category, not a reason to care.

Most of the remaining copy is labels: model names, ENCRYPT, PROVE, VERIFY, CHAT, AGENTS, SDK/API. There's no proof density, no attributed evidence, and the emotional registration lives entirely in that one privacy line. The page reads as a visual system with captions rather than written argument.

Category reads fast, audience and reason-why stay fuzzy.

The category signal is the page's strongest clarity asset. "Switch between 30+ providers through a single unified interface" tells me what Scroll is quickly, and the provider logos reinforce it. A technical visitor will understand the product within ten seconds.

The identity signal is weaker. The content clearly points at developers (SDK/API, BYOM, model names), but no one is named, so I have to infer that I'm the audience. Differentiation is partial: the ZK privacy angle is a real wedge, but it's expressed as labels (ENCRYPT, PROVE, VERIFY) rather than a stated advantage over calling providers directly or using an existing router.

Jargon load is manageable for the intended evaluator: a developer will parse ZK proof, BYOM, and the model names fine. Cognitive sequencing is the problem. The page leads with a diagram and interactive toys described by their loading states, and the clearest value lines arrive late and scattered, so the order doesn't match the questions a visitor actually asks.

Strong privacy beat, almost no customer or stakes.

The page delivers a handful of beats well through its diagram and labels. The functional job is clear: switch between 30+ providers through one interface. The privacy differentiation ("no prompts stored, no keys exposed, provider blind") carries both a functional and an insight-adjacent beat. The plan is partially visible through the ENCRYPT, PROVE, VERIFY, ROUTE sequence, though it reads as labels rather than steps I can follow.

Where the story goes quiet is the human side. I never learn who the visitor is: a developer, an agent builder, an enterprise buyer worried about data leakage. The emotional and social jobs are implied by the privacy promise but never named. There's no empathy beat, no "we know you're tired of managing ten API keys and worrying where your prompts go." And there's no stakes beat at all: nothing about what persists if you keep routing calls the leaky way.

Authority is absent in the proof sense. There are model logos and provider names, but no customers, no numbers, no attributed evidence that Scroll's proofs actually work. The architecture is detailed; the argument for why I should trust and adopt it is thin.

The functional job is clear through "switch between 30+ providers," but the visitor's role, emotional job, and social job are never named, so I have to assume I'm the audience.

The privacy triad implies the problem of exposed prompts and keys, but no obstacle is stated as a recognizable pain the customer would nod at.

There's an insight wedge in "provider blind" and ZK proofs, but no empathy and no attributed authority or proof that the system works.

The ENCRYPT, PROVE, VERIFY, ROUTE diagram shows a mechanism and the product trio hints at how adoption works, but it reads as labels more than steps.

"One key. Every provider. No prompts stored." hints at a better after-state, but there's no identity-level transformation and no stakes for inaction.

Your primary CTAs sit above the fold and are visible: "Check it out" and "ZK API keys" are both present in the hero, with a dark primary button giving a clear visual hierarchy. That's a reasonable setup.

The weakness is specificity and motivation. "Check it out" tells me almost nothing about what happens next, and "Jointhewaitlist" and "Download" later on pull in different directions. There's no reason to act now rather than later: no urgency, no articulation of what I gain today. The ZK API keys CTA is the most concrete, but it reads as a label more than an invitation.

Friction is moderate and uneven. A waitlist implies the product isn't fully available, which conflicts with "Check it out" and "Download" suggesting it is. That mix of signals creates small decision friction, and none of the CTAs are tied back to the privacy story that is the page's best argument.

How we score

Clarity Score v7 reads each company's live homepage, words and visuals together, and scores four dimensions on a 0 to 100 scale.

What does Clarity Score measure?

Four dimensions, weighted into the overall score:

Narrative Completeness 30% · does the page tell a complete story, from who it is for to what changes for them.
Message Quality 30% · is the writing actually good: specific, distinctive, and worth reading.
Signal Clarity 20% · does a stranger get what this is and who it is for, fast.
CTA 20% · does it say what to do next, with a reason to do it.

How often is the board updated?

Every company is re-scored monthly. The Change column shows movement since the prior reading; the board baselined on Clarity Score v7 in June 2026.

What does Clarity Score not measure?

It measures how clearly a homepage communicates. It does not measure product quality, funding, traction, or whether a company is winning.