# Content Quality Audit — Quire Guide (https://quire.io/guide/) Method: Live fetch of the guide homepage plus direct inspection of the source-of-truth Jekyll content (`_docs/en/*.md`, 139 English pages) and layout/include files (`_layouts/single.html`, `_includes/_seo.html`, `_includes/footer.html`, `_includes/ai_translated_banner.html`) that generate the live HTML. Sampled ~12 pages across categories (homepage, overview, project, task list, table/board views, formula, security, troubleshooting, account, mobile) plus a German translation spot-check. ## Summary Score: 84/100 The guide is unusually strong for a help-center property: near-universal adherence to a documented intro/H2/H3/FAQ convention, full frontmatter + visible FAQ coverage on every one of 139 English pages, solid site-wide Organization/trust schema, and an explicit AI-translation disclosure banner. The main deductions are thin-content risk on a meaningful minority of short procedural pages, a stale/missing freshness signal (`last_modified_at` is defined in the template but never populated, so ~86% of pages carry a static 2019-08-23 publish date despite evidence of recent substantive rewrites), and the total absence of author/editor bylines or credentialing anywhere in the corpus. ## What Works - **Structural convention compliance is excellent.** Every one of the 139 EN pages sampled and spot-checked has: (1) an intro paragraph between frontmatter and the first H2, (2) H2s used for major sections, H3s used as topic labels rather than questions (e.g. `### In the Workspace`, `### Restrictions for Organization Admins`), and (3) a `### Frequently Asked Questions` section at the bottom. This checks out via `grep -L "Frequently Asked Questions"` returning zero files across `_docs/en/`. - **AI-citation readiness is high.** All 139 pages carry a frontmatter `faq:` block whose entries are mirrored (in shorter form) as visible on-page FAQ answers — a good split between long-form SEO answers and short, self-contained, quotable answers. Numbered steps are used consistently for sequential UI actions; bold UI labels and `` tags make instructions machine-parseable. - **Trust/authority signals at the site level are strong.** `_includes/_seo.html` emits Organization schema with founder name, founding date, parent company, and `sameAs` links to G2, Trustpilot, Capterra, GetApp, Software Advice, and Wikipedia — solid third-party-recognition signals for E-E-A-T. The footer exposes a real company name (Potix Corporation), contact email addresses (`info@quire.io`, `feedback@quire.io`), links to `/security`, `/privacy`, `/terms`, and a persistent in-page "send feedback" widget on every guide article. - **Transparent AI-translation disclosure.** `_includes/ai_translated_banner.html` shows an explicit "AI translated" banner with a link back to the English source on every non-English page — exactly the kind of disclosure Sept 2025 QRG rewards rather than penalizes, and it avoids passing off machine translation as original-authored content. - **Translation parity is nearly complete.** German has 137 of 139 English pages (only `templates-03-okrs` and `templates-04-gtd` missing); spot-checked DE translation of `tasklist-01-add-tasks.md` preserves the same H2/H3/FAQ structure and heading anchors as the English source, suggesting the pipeline maintains structural fidelity across languages, not just prose translation. - **No literal duplicate content.** Templated sentence patterns recur across topically-adjacent pages (e.g. `leave-organization` vs `leave-project`), but each pair covers materially different UI flows/restrictions — this is consistent procedural documentation style, not copy-paste duplication. ## Findings ### 1. Thin content on a meaningful subset of pages **Severity: Medium** Body word counts (frontmatter excluded) across the 139 EN pages range from 140 to 4,706 words (mean 751, median 675). 35 pages (~25% of the corpus) fall under 400 words, and 8 of those are under 200 words, e.g. `troubleshooting-02-compatible-browsers.md` (140 words), `tasklistad-14-print.md` (166), `tasklistad-13-activity-log.md` (169), `tasklistad-04-zoom.md` (181). These are legitimate single-feature help articles, not padded content, but at 140–250 words several offer minimal topical coverage beyond a 2-step instruction and a 2-question FAQ, which is thin relative to the "product page" floor (300+) in the QRG rubric and could read as low-value to both users and crawlers comparing against competing help centers with more thorough coverage (troubleshooting steps, screenshots per browser, common error scenarios, etc.). **Recommendation:** For the sub-200-word tier specifically, expand with genuinely useful detail (e.g. common error messages, what happens if the feature doesn't work as expected, links to 2–3 related tasks) rather than word-count padding. Do not force pages that are inherently simple (e.g. `compatible-browsers`) past a natural length just to hit a threshold — QRG explicitly treats word count as a coverage floor, not a target. ### 2. No freshness/last-modified signal despite active content rewrites **Severity: Medium** `_layouts/single.html` and `_includes/_seo.html` are both wired to render `page.last_modified_at` (as visible "Updated" text and as `dateModified` / `article:modified_time` schema) but no page in `_docs/en/` sets this field — a grep for `last_modified_at:` or `updated:` across all 139 files returns zero matches. Meanwhile 120 of 139 pages (86%) carry the identical static `date: 2019-08-23`, even though git history shows the FAQ frontmatter and rewritten body copy on pages like `tasklist-01-add-tasks.md` were substantively edited as recently as May 2026. This means both users and Google see a 2019 publish date with no indication the content was recently and materially updated — a missed freshness signal that actively undersells the site's actual content-maintenance cadence. **Recommendation:** Populate `last_modified_at` (via `jekyll-last-modified-at` plugin or a manual field set during the editorial/FAQ-rewrite pass) so the schema `dateModified` and the visible "Updated" date reflect real edit history. This is a low-effort, high-value fix given the template already fully supports it. ### 3. No author/editor bylines or credentialing anywhere in the corpus **Severity: Low** `index.md` explicitly sets `author_profile: false`, and no page in `_docs/en/` has an `author:` field. There is no named author, support-team credential, or "reviewed by" signal on any guide page — only the site-level Organization schema (founder, parent company) provides any authorship context. For a SaaS help center this is common and not disqualifying, but Sept 2025 QRG increasingly rewards visible expertise signals even for practical/procedural content. **Recommendation:** Consider a lightweight, site-wide "Written and maintained by the Quire Support Team" byline/schema (Organization-as-author is acceptable under schema.org `Article.author`) rather than per-writer bylines, to close this gap without overhauling editorial workflow. ### 4. Homepage does not follow the page's own H2/intro convention **Severity: Low** `index.md` opens with `### Welcome to Quire's ultimate guide!` (an H3) followed by the intro paragraph, then `## Highlights` — i.e., the heading order is H3-before-H2, which inverts the documented convention (intro paragraph should precede the first H2, and H3 should be subordinate to an H2, not precede it). The homepage also has no FAQ section, unlike every other guide page. **Recommendation:** Either drop the H3 "Welcome" heading and let the existing paragraph serve as the plain intro (matching every other page in the corpus), or promote it to context appropriate to a landing page. FAQ omission on the homepage is reasonable (it's a hub page, not a topic page) but worth an explicit exception note in `GUIDE-FORMAT.md` so it doesn't get "fixed" inconsistently later. ### 5. Formulaic sentence templates repeated across near-identical procedural pairs **Severity: Info** Pages covering structurally parallel actions (e.g. `org-06-leave-org.md` vs `project-12-leave-project.md`) reuse near-identical sentence templates ("Click the dropdown menu icon next to X, select Edit members, scroll to the bottom, click Leave X..."). This is legitimate given the underlying UI flows are genuinely parallel, not duplicate content, but at scale (139 pages, likely more such pairs beyond the sample) this pattern is worth a periodic manual scan to ensure it doesn't drift into truly interchangeable, low-differentiation copy — one of the "repetitive structure across pages" markers the Sept 2025 QRG flags for low-quality AI-assisted content when it's not paired with page-specific specificity (which, in the sampled pairs, it currently is: each page correctly reflects distinct restrictions/flows for orgs vs. projects). **Recommendation:** No action required now; flag for periodic review if new templated page batches are generated (e.g. via the `seo-programmatic` pattern) without a sameness/uniqueness check. ## Not Assessed / Out of Scope for This Pass - Live-render fetch of https://quire.io/guide/ pages beyond the homepage returned truncated content via the automated fetch tool (503-byte response vs. 36.6KB via direct `curl`), so this audit relied on the checked-out Jekyll source (`_docs/en/*.md`, `_layouts/`, `_includes/`) as the authoritative representation of the deployed HTML rather than crawling the live site page-by-page. Source-vs-deployed parity was not independently verified beyond confirming header/response codes and one successful `curl` fetch of the homepage. - Full 13-language parity was not checked beyond a German file-count diff and one German page structural spot-check.