The Scribe-X Homepage Face-Off
Two renditions of the new homepage, scored section by section through six lenses: how an LLM reads it, how it holds up under ADA/WCAG, how the ideal client (an FQHC buying committee) experiences it, how it converts, how well it explains the model visually, and how faithfully it executes the locked MMF (V4, July 6). Every number below comes from a live audit of both pages against the framework, not from taste.
A split decision ... and that's the real finding.
Scored head-to-head, the pages finish within a tenth of a point of each other, but they win in opposite categories. Greg's page wins on message and strategy: it carries the locked positioning spine, mirrors the full buying committee, and tells the truth about what's live versus coming. Oliver's page wins on execution mechanics: cleaner semantics, a dedicated security section, an embedded lead-capture quiz, and the best single trust device on either page (the five vendor questions).
The recommendation: don't pick a winner. Build the composite.
The MQL machine Jason needs is Greg's narrative with four of Oliver's organs transplanted in: the security section, the embedded quiz, the five vendor questions, and the single-message hero discipline. Section-by-section tally: Greg wins 8, Oliver wins 5, 2 ties. Across the six lenses, Greg takes ICP, MMF fidelity, and visual explanation; Oliver takes CRO and ADA; LLM readability is a near-tie. The blueprint is at the bottom of this page.
Oliver's greyscale-logo claim: your instinct is right.
Greyscaling the nav logo does nothing for ADA compliance.
WCAG (the standard ADA web cases are measured against) is explicit: text that is part of a logo or brand name has no contrast requirement. That's Success Criterion 1.4.3, and the logotype exemption carries into the non-text contrast rule (1.4.11) as well. A full-color Scribe-X logo in the nav is exactly as compliant as a greyscale one. Making it grey is an aesthetic choice, and it's fine as an aesthetic choice on his dark header ... but it should be argued as design taste, not compliance.
What actually moves ADA risk on these two pages: body-text contrast (both pages fail today ... details in the technical appendix), a <main> landmark and skip link (both missing on Greg's, skip link missing on Oliver's), and Oliver's scroll-hijacked problem section, which needs a prefers-reduced-motion fallback. If the team wants to spend ADA effort, spend it there. This is just truth.
Section by section: what earns its keep in the scroll.
Each round shows what both pages do, who wins, and whether the section earns its scroll space in the composite. Verdicts: Keep stays as-is, Graft gets transplanted into the other page, Compress earns a smaller footprint, Cut doesn't earn the space.
Navigation
Greg winsKeepHero
SplitCompress (pick one)The Problem Story
Greg winsKeep Greg's · compress Oliver'sThird-Party Evidence
Oliver winsGraft into Greg'sPlatform Tiers (Pulse / Assist / Live)
Greg winsKeep Greg's + graft Oliver's calculatorWhole-Journey Coverage
Greg winsKeepSecurity & IT
Oliver winsGraft into Greg's ... biggest single transplantCompetitive Alternatives
Oliver winsGraft the 5 questionsOutcomes / Proof
TieBoth need verified numbersTestimonials
Greg winsCompressWho This Is For (Buying Committee)
Greg wins by forfeitKeepLead Capture / Front Door
Oliver winsGraft the embedCost of Standing Still
Greg wins by forfeitKeepFAQ
Greg wins on substanceMerge the two setsFinal CTA & Footer
TieKeep eitherThe five heroes, ranked.
For the Jason + Ethan vote. Ranked on spine fidelity, specificity, ICP resonance, and LLM quotability.
- D · "The AI scribe with a human at the center." ... It IS the locked positioning spine, verbatim. Ownable (no competitor can say it), differentiating in six words, and the phrase an LLM should learn to associate with Scribe-X. Oliver voted D in June. This is the one.
- E · "Not just the note. The whole visit." ... The whole-journey belief, punchy, category-reframing. Best runner-up; strong section header inside the page even if it loses the hero.
- A · "Your AI scribe clocks out at the note. The work doesn't." ... The best problem-voice line of the five, but it assumes the visitor already runs an AI scribe. Perfect for the failed-pilot segment: use it in ads and LinkedIn, not as the front door for everyone.
- C · "Give your providers their day back." ... Warm, human, outcome-led ... and any scribe vendor could say it. Keep it as the final CTA header (it's already there).
- B · "Patients, not paperwork." ... A credo, not a headline. It's already the page title and the footer sign-off, which is exactly where it belongs. As a lone H1 it gives a first-time visitor (and an LLM) too little to work with.
Visual explanation: the hero diagram I should have scored the first time.
Fair callout from Greg, and he's right. My first pass scored "visual" as cinematic polish ... photography, mood, single-message discipline ... and on that reading I gave Oliver the hero. But that conflates looking good with explaining the model, and those are different jobs. Greg's hero carries an actual information graphic that teaches the entire positioning in one glance, above the fold, before a word of body copy. That's the harder and more valuable thing a B2B homepage can do, and I under-weighted it. Here's the diagram, rebuilt so the point is concrete:
Rebuilt from Greg's live hero (captured 2026-07-16). In the real thing the bar is a color gradient light→dark and the handle animates in.
Why this one graphic is the strongest explanatory asset on either page
Read it once and you've absorbed the entire pitch: the category name (CIP), the three solutions as positions on a bar, their availability (Pulse grayed + COMING, Assist and Live live), the per-provider toggle (the draggable handle), the whole eight-stage journey, the wedge (only THE NOTE lights up ... the one stage AI-only covers, with seven others sitting untouched), and the spine itself rendered as the axis: AI only on the left, human at the center on the right. That is MMF §10 ... "the category-defining asset" ... turned into a picture. It does in one glance what Oliver's page needs several scrolling sections to say.
The honest limits: it's dense, so it rewards a second look more than a glance, and at mobile width it stacks and the journey labels shrink to the point where the ADA lens already flagged them for low contrast. Worth simplifying the small-screen version. But as a desktop above-the-fold explainer, nothing on Oliver's page matches it.
| Visual asset | Oliver | Greg |
|---|---|---|
| Hero visual | Cinematic Full-bleed clinical photography, gradient accent, mood. Beautiful, but decorative ... it doesn't explain the model. | Explanatory The CIP toggle-bar diagram above. Teaches the whole positioning before any scrolling. |
| Problem, made visual | Strong The 400vh scroll-jacked "work of one visit" grid (AI handles the note; everything else stays on providers). Memorable ... but motion-heavy and hidden on mobile. | Strong The 7-row AI-only vs AI+human table maps 1:1 to the MMF shift table. Static, scannable, crawler-safe. |
| The mechanism, made interactive | Best-in-class The mix-builder ("add your providers, see the mix") is a real interactive explainer. Worth grafting. | Present The toggle bar + the AI/human overlap visual carry it, though not as an interactive widget. |
| Journey coverage, made visual | Clean Three intelligence-layer cards (before / during / after). | Fuller Seven journey stages with live status labels, echoed in the hero rail. |
Net: Oliver wins the screen on craft; Greg wins it on explanation. For a CMIO who has to grasp "how is this different from the AI we already have" in five seconds, the diagram does more work than the photograph. Greg 9.0, Oliver 7.5. And it reframes the Hero round below: my "Oliver wins the screen" note stands only on aesthetics ... on teaching the model, the diagram is Greg's decisive edge.
MMF fidelity: does the page execute the framework Jason signed off on?
This is the lens that turns "Greg's taste vs Oliver's taste" into something objective. The rubric isn't opinion ... it's the locked Magnetic Messaging Framework (V4, July 6, 2026): the spine, the toggle-bar mechanism, the RHA naming decision, the buying committee, the credo, and the Language Library's do/don't list. Each page is scored on whether it carries those load-bearing elements. Greg wins 8.5 to 5.5, the widest gap on the board, and the reason is structural: the framework's central sentence exists on only one of these two pages.
| MMF element (V4) | Oliver | Greg |
|---|---|---|
| The spine ... "a human at the center, not at the edge" | Absent The exact phrase the whole Phase 1 hangs on never appears. Hero is "More than an AI medical scribe," which concedes the category instead of reframing it. | Verbatim Hero Option D is the spine word-for-word; subhead names the mechanism (AI volume, human judgment). |
| Category name ... Clinical Intelligence Platform (CIP) | Present Used as the hero eyebrow and throughout. | Present "THE SOLUTION · The Clinical Intelligence Platform." |
| The mechanism ... the toggle bar, set per provider | Present "One platform. Set per provider," with a mix-builder and the AI↔human axis. | Present + fuller "How much human does each provider need?" with who-it-fits and what-you-keep. |
| Three solutions ... Pulse / Assist / Live (not the retired tier names) | Named right Uses the correct solution names, not Essentials/Professional/Enterprise. | Named right Same. |
| Pulse shown grayed / COMING, never promoted | Violated Presents Pulse as live and tags it "where most organizations start." MMF is explicit: Pulse stays grayed until operationally ready, and Assist is the start-here tier. | Correct Pulse labeled COMING; "MOST PRACTICES START HERE" sits on Assist, exactly per §11. |
| People term ... "Remote Healthcare Assistant (RHA)" | Correct Uses "Remote Healthcare Assistant" in body and meta. | Fixed 2026-07-16 Was inconsistent (heroes said "clinical specialist"); all five heroes now read "US-based Remote Healthcare Assistant," matching the body and the July 6 naming decision. Consistent end to end. |
| The buying committee ... CMO champion, CMIO bridge, CIO risk (§6) | Absent No persona section; the committee is implied, never named. | Full Three persona cards matching the three clinical doors, each with an inner-monologue quote. |
| The credo ... "Patients, Not Paperwork" | Absent The locked credo never appears. | Present Page title, Option B hero, and the final CTA. |
| The Shift ... AI-only vs AI + Human Intelligence (§4) | Strong "The industry already ran the AI-only experiment," with JAMA. Best framing of the shift on either page. | Strong The 7-row "what matters most" table maps almost 1:1 to the MMF's shift table. |
| The villain ... "Set-and-Forget AI" (the idea, not AI itself) | Implied Frames AI-only as insufficient but never names the villain. | Implied Same ... neither page names it outright. Room for both to sharpen. |
| The whole patient journey ... before / during / after | Present Three intelligence layers. | Present Seven journey stages with status labels. |
| Cost of doing nothing (§9) | Absent Inaction is never priced. | Present Six compounding costs section. |
Why the gap is structural, not cosmetic
Three of the MMF's load-bearing pillars ... the spine sentence, the buying committee, and the credo ... are simply not on Oliver's page. Those aren't finishing touches; they're the framework's answer to "who are we, who buys, and what do we stand for." A page can be beautifully built and still not be executing the strategy, and that's the honest read here: Oliver's is the stronger craft, Greg's is the stronger framework execution.
The scoring cut both ways, which is how you know it's the framework talking and not a thumb on the scale: Greg's page carried a real MMF violation of its own ... all five heroes used "US-based clinical specialist," the exact term Jason and Greg retired on July 6 for "Remote Healthcare Assistant." That's now fixed (2026-07-16): the source was find-replaced across all five heroes, so the copy is consistent end to end and Greg's fidelity score rose from 8.5 to 9.5 on the revote. One deploy step remains to push the fix to his live draft URL (it targets Scribe-X's Cloudflare account and needs a go-ahead).
The audit receipts.
| Check | Oliver (CIP homepage) | Greg (web draft) |
|---|---|---|
| H1 discipline | 1 H1, clean H2/H3 hierarchy | 5 H1s (the stacked heroes ... resolves to 1 after the vote) |
| Landmarks | main present; header/nav/footer correct | No main landmark; header/nav/footer correct |
| Skip link | Missing | Missing |
| Meta description | Present, strong (names CIP + US-based human + whole visit) | Missing (draft state) |
| JSON-LD structured data | None | None ... both should ship FAQPage + Organization schema |
| Open Graph tags | 0 | 9 |
| Image alt text | 31 images, all with descriptive alts (incl. product shots) | 22 images, 10 empty-alt (decorative, acceptable) |
| Contrast fails (solid backgrounds, WCAG AA) | 11 ... worst: amber stats 1.76:1, lime stats 2.3:1, slate labels 2.45:1 | 13 ... worst: amber COMING badge 1.76:1, light-blue labels 1.96:1, CTA buttons marginal at 4.02:1 |
| Motion risk | 400vh scroll-hijacked section; no reduced-motion fallback found; hidden on mobile | Scroll-reveal fades (50 blocks start at opacity 0); milder, still wants a reduced-motion guard |
| Content without JavaScript | Text renders (reveal states applied after hydration) | 50 ".reveal" blocks invisible without JS (CSS default opacity 0) |
| Stats served to crawlers | Broken: "0.0–3.8", "0–62%", "Up to 0%" (count-up initial zeros in HTML) | Correct: real values ("10%", "1 in 3") in raw HTML |
| Server-rendered text volume | ~12.7K chars | ~21.7K chars |
| Stack | Next.js, static export | Hand-built static HTML (Tailwind) |
Flags to raise with Oliver / Jason (beyond the scoring)
- Pulse truth-labeling: Oliver's page presents Pulse as available today; the agreed state is COMING. Whatever ships must match the toggle-bar reality.
- "Start a free trial" on Greg's front-door section is an undefined offer ... define it with Jason or pull the button.
- The positioning spine is absent from Oliver's page. "A human at the center" ... the phrase the whole Phase 1 hangs on ... never appears in his copy. That's the deepest fix, not a cosmetic one. (See the MMF fidelity lens above for the full framework gap.)
- Greg's heroes used a retired term ... RESOLVED 2026-07-16. All five hero options said "US-based clinical specialist"; the July 6 decision replaced it with "Remote Healthcare Assistant." Find-replaced across all five in the source. Live draft URL still needs the one redeploy (targets Scribe-X's CF account).
- Access gate leak: Oliver's Cloudflare Access gate doesn't cover deployment-hash URLs (this audit read the page through one, no login). If the gate matters, it should cover *.scribex-cip-homepage.pages.dev.
- Both pages need before launch: verified outcome numbers (Justin), JSON-LD (FAQPage + Organization), contrast fixes listed above, skip links, and the standing GTM + HubSpot tracking block once the page is public-facing.
The composite page that wins all four lenses.
1. Hero ... Option D copy ("The AI scribe with a human at the center") over Greg's CIP toggle-bar diagram (the single best explainer on either page), with Oliver's photographic polish on the treatment and trust logos immediately below.
2. Problem ... Greg's three committee quotes + the 7-row comparison table. Oliver's cinematic grid compressed to one screen, if kept at all.
3. Evidence ... Oliver's "The industry already ran the AI-only experiment" section, verbatim placement.
4. Platform ... Greg's toggle-bar depth (who it fits / what you keep / honest status labels) + Oliver's mix-builder calculator grafted in.
5. Security & IT ... Oliver's section, transplanted whole. The CIO's card, answered.
6. Alternatives ... Oliver's appeal/catch framing + the five vendor questions.
7. Who this is for ... Greg's three persona cards, untouched.
8. Proof ... verified cohort ranges (post-Justin), scoped per solution, real values server-rendered. Three tight testimonials.
9. Front door ... Greg's section with Oliver's quiz embedded in-page.
10. Cost of standing still ... Greg's six costs, untouched.
11. FAQ ... Greg's buyer-voiced question set, Oliver's answer tightness, no AI-chat conceit.
12. Close ... "Patients, not paperwork. Give your providers their day back." Dual CTA. Done.
The composite isn't a compromise ... it's the page that scores highest on all five lenses at once: Oliver's craft and conversion mechanics carrying Greg's framework execution. When this goes to Jason, lead with the MMF fidelity lens. It reframes the whole conversation from "whose page is better" to "which page executes the strategy you already signed off on," and it's the lens where the answer is least about taste.