The conversion path needs repair before growth work compounds.
A focused public-site review of www.gwenth.com. The site has a sound crawlable baseline and clear positioning; the next priority is making the highest-intent path reliably actionable, then removing friction from the desktop story.
Gwenth clearly presents an AI sales execution platform for founders and lean B2B software teams, with product vocabulary spanning Find, Connect, Sell, Manage and a Command layer. Public technical signals observed in this pass are broadly constructive. But the visible hero “Book a demo” control did not produce an observed outcome in mobile or manual desktop testing, while a multi-viewport-height blank desktop narrative gap breaks continuity before a comparison section.
Fix those P1 issues first. On mobile, restore a persistent route to key pages and conversion. Then validate the detailed technical, accessibility and AI-search work below rather than treating the observed baseline as a completed audit.
02 / PRIORITY FINDINGS
What to fix, in order.
Labels distinguish observed defects from recommendations and follow-up validation. Severity reflects likely impact on acquisition and comprehension, not an assessment of the application or all site routes.
P1VERIFIED DEFECT
Mobile hero “Book a demo” CTA is inert
Observed impact
At 390 × 844, a visible 350 × 53 px hero button accepted a click but did not change URL or scroll position, or open a visible dialog after 700 ms. A manual desktop click also had no visible outcome. This makes the headline conversion action appear actionable without starting a booking path.
Recommended fix
Wire the CTA to a real scheduler or lead form, or a visible keyboard-accessible modal. Make the resulting destination and expected next step explicit.
Effort
Small to medium
Verification
End-to-end test: click/tap CTA, assert the form or scheduler becomes visible and operable by keyboard; repeat at mobile and desktop breakpoints.
P1VERIFIED DEFECT
Desktop narrative has a large scroll dead-zone
Observed impact
After the Command product-view section, desktop capture showed blank space spanning several viewport heights before “One founder juggling tools…”. The surrounding #narrative section was explicitly lg:h-[560vh] and measured 3,231 px at the reviewed desktop viewport.
Recommended fix
Reduce or remove the fixed/minimum height, unless a scroll-driven state continuously earns the distance. If it is intentional, ensure meaningful visual or textual progression throughout.
Effort
Medium
Verification
Record a normal desktop scroll through the full narrative and confirm continuity, loading states, and no unexplained blank gap.
P2VERIFIED DEFECT
Mobile header hides primary navigation without a menu
Observed impact
At 390 px, the header visibly contained the brand mark and “Blog” only. System, How it works, Workflow, Why Gwenth, Sign in and header Book a demo were hidden/zero-sized; no menu trigger was found. Combined with the inert hero CTA, there was no observed persistent mobile conversion route.
Recommended fix
Add an accessible menu trigger with labeled state, focus management and Escape behavior, or retain a compact working CTA. Preserve usable tap targets and focus order.
Effort
Medium
Verification
Test keyboard and touch navigation at narrow widths, including open/close, focus return, every visible link and the conversion route.
P2RECOMMENDATION
Give the lower CTA a differentiated next step
Observed impact
The lower section repeats “Book a demo” and “See how it works.” This is a conversion-friction observation, not an accessibility failure.
Recommended fix
After booking works, make the secondary path concrete—for example, a short product tour—and add nearby proof or expectation-setting copy.
Effort
Small
Verification
Instrument CTA choice and downstream completion; compare clarity and completion after sufficient traffic.
P3FOLLOW-UP VALIDATION
Review dense product illustrations for access and responsive comprehension
Observed impact
The five-surface cards and Command demo are information-dense. No horizontal overflow or JavaScript console errors were observed at 390 px, but this is not an accessibility conformance result.
Recommended fix
Validate font scaling, touch-target size, keyboard traversal, alt/semantic equivalents, and whether meaningful product states are available as text.
Effort
Medium
Verification
Run real-device and keyboard checks with zoom/text scaling, then remediate any identified issues.
03 / EXPERIENCE & CONVERSION
Make the route through the story obvious.
Navigation & hierarchy
Keep the brand, primary navigation and conversion path available at every breakpoint. Preserve semantic landmark structure, a logical heading hierarchy, visible keyboard focus and a skip link. Test the mobile menu rather than assuming a hidden desktop pattern is usable.
CTA system
Use one dependable primary action with a clear destination. Pair secondary actions with an outcome, not a vague label. Place proof and timing expectations beside high-intent CTAs; measure the entire path from click to qualified submission.
Responsive/accessibility follow-up
The captured mobile page had no horizontal overflow and one rendered H1. Do not create a duplicate-H1 defect from raw markup. Validate real-device behavior, scaling, touch targets, focus order, dialog behavior and meaningful non-visual equivalents before claiming compliance.
04 / DISCOVERABILITY
Keep the crawlable baseline; validate the claims behind it.
Title, description, canonical, Open Graph and Twitter card observed.
robots.txt and sitemap are available; a 17/17 sitemap URL crawl sample returned HTTP 200.
Home HTML contains two JSON-LD scripts.
NEXT VALIDATION / REMEDIATION
Validate canonical consistency across indexable routes; inspect sitemap freshness and robots directives in production; run structured-data testing and fix any errors found. Validate rendered metadata and social previews. PageSpeed was not measured: the public API returned 429 RESOURCE_EXHAUSTED, a shared-quota limit—not evidence of a performance failure.
GEO / AI-search
OBSERVED STRENGTHS
Clear positioning and publicly stated product vocabulary.
Robots policy permits general crawling and explicitly permits OAI-SearchBot.
Public content includes About, Blog and vertical pages.
IMPLEMENT & VALIDATE
Make clear entity, About and contact evidence easy to find. Support factual and product claims with citations; write answer-first pages with descriptive headings; validate structured data; show editorial authorship; and publish original expert content. llms.txt is optional and experimental—not a magic ranking factor or a crawl substitute.
05 / 30–60–90 DAY PLAN
Sequence repair, validation and learning.
0–30
Restore paths
Repair and test hero booking; restore mobile navigation/CTA; remove or justify the desktop dead-zone. Add CTA funnel events and consent-aware analytics.
31–60
Prove baseline
Run real-device, keyboard and text-scaling checks. Validate structured data and rendered metadata. Review canonical, robots and sitemap behavior across intended indexable pages.
61–90
Build authority
Publish answer-first expert content with authorship and cited claims. Improve entity/about/contact evidence. Review CTA conversion data and iterate on the lower-funnel path.
Measurement plan
Connect and monitor Google Search Console and Bing Webmaster Tools. Use consent-aware analytics to track CTA impressions, clicks, booking-form views, submissions and qualified outcomes. Review server logs for crawler behavior. Assess field Core Web Vitals only after sufficient eligible traffic; no CrUX or PageSpeed result was collected in this pass.
06 / METHOD, LIMITS & PROVENANCE
A bounded public review, not a certification.
Method
Anonymous public desktop and 390 × 844 mobile-browser rendering; CTA interaction test; HTML and HTTP-header review; robots and sitemap review; sitemap crawl sample.
Limits
No real-device pass, logged-in flow test, structured-data validator, CrUX or PageSpeed measurement. No claim is made about unaudited routes, application security, accessibility conformance, rich-result eligibility, performance, or conversion outcomes.
Evidence appendix
Primary provenance: source/observed-evidence.md, collected from 2026-07-12T15:24:49Z onward for the canonical public host. Supporting captured public artifacts remain in source/. This report intentionally distinguishes what was observed from recommendations and work still requiring validation.