# AI Search Readiness (GEO) Audit — quire.io/guide/

Audited: 2026-07-02
Scope: robots.txt, llms.txt, sitemap.xml, rendered HTML of representative page (`/guide/add-task/`), and static source of all 139 English `_docs/en` pages (structure/FAQ analysis extrapolates to the other 12 translated locales, which share the same Jekyll templates and schema partials).

## GEO Readiness Score: 77 / 100

| Dimension | Weight | Score | Weighted |
|---|---|---|---|
| Citability | 25% | 78 | 19.5 |
| Structural Readability | 20% | 85 | 17.0 |
| Multi-Modal Content | 15% | 55 | 8.25 |
| Authority & Brand Signals | 20% | 70 | 14.0 |
| Technical Accessibility | 20% | 90 | 18.0 |
| **Total** | | | **76.75 ≈ 77** |

Note: No DataForSEO MCP tools were available in this environment, so platform scores below are estimated from technical/content readiness signals (crawler access, schema validity, passage structure), not from live ChatGPT/Perplexity query testing. Treat them as directional, not measured.

## AI Crawler Access Status (robots.txt)

`https://quire.io/robots.txt` uses a blanket `User-agent: * / Allow: /` with only `/api`, `/oauth`, `/qa`,`/qb`,`/qr`,`/qs`, and `/r/*` admin/auth paths disallowed — none of which live under `/guide/`. Because there is no explicit rule for most AI bots, they inherit the wildcard `Allow: /`.

| Crawler | Status | Notes |
|---|---|---|
| GPTBot | Allowed | Inherits wildcard `Allow: /` |
| OAI-SearchBot | Allowed | Inherits wildcard `Allow: /` |
| ClaudeBot | Allowed | Explicit block with `Crawl-delay: 5`, `Allow: /` |
| PerplexityBot | Allowed | Inherits wildcard `Allow: /` |
| Google-Extended | Allowed | Inherits wildcard `Allow: /` (no dedicated rule found) |
| Bingbot / BingPreview | Allowed | Inherits wildcard `Allow: /` |
| CCBot | Allowed | Not explicitly blocked — inherits wildcard. If Potix wants to permit AI-search crawling but opt out of training-only scraping, CCBot and anthropic-ai should get their own `Disallow: /` blocks (recommended optional hardening, not a defect) |
| Bytespider | Allowed | Explicit block with `Crawl-delay: 10`, `Allow: /` |

`Sitemap: https://quire.io/sitemap.xml` is declared and resolves (200) to a sitemap index referencing `sitemap-main.xml`, `blog/sitemap.xml`, and `guide/sitemap.xml` — all machine-discoverable.

**Verdict: no crawler-access blockers.** Every major AI-search bot can reach and crawl the guide content today.

## llms.txt Status: Missing (confirmed)

- `https://quire.io/llms.txt` → 404
- `https://quire.io/guide/llms.txt` → 404
- No RSL 1.0 licensing signal found anywhere in checked pages

Given the guide's scale (139 English pages × 13 locales ≈ 1,800+ URLs, already covered by a clean `sitemap.xml` index, permissive `robots.txt`, canonical tags, and full hreflang alternate links on every page), an `llms.txt` is a **nice-to-have, not a blocker**. It would not unlock crawl access that's currently missing — everything is already SSR and crawlable — but it would give LLM ingestion pipelines a curated, deduplicated entry point (one canonical URL per topic instead of 13 locale variants) and a place to assert licensing/usage terms. Priority: Low-Medium, mainly valuable once the localization footprint grows further or if Potix wants to control which locale is treated as canonical for training ingestion.

## Passage-Level Citability

Checked: does the guide's FAQ format (frontmatter `faq:` block for SEO + visible "Frequently Asked Questions" section per `GUIDE-FORMAT.md`) actually produce citable passages?

**What works well:**
- All 139/139 English pages have both a frontmatter `faq:` block and a rendered visible FAQ section — 100% coverage, no gaps.
- FAQ content is emitted server-side (present in raw curl HTML with no JS execution), so every AI crawler sees it without rendering.
- Visible FAQ answers follow the house style (1-2 sentences, self-contained, no em dashes per `GUIDE-FORMAT.md`) — genuinely extractable without surrounding context.
- Each page also carries a valid `FAQPage` JSON-LD block whose `Question`/`Answer` pairs match the visible FAQ text verbatim, reinforcing the same passages for engines that parse structured data.
- Every page has an intro paragraph and each H2 section leads with direct, procedural (often numbered) instructions — good for step-based citation ("How do I add a subtask in Quire?").

**Gap found — most FAQ answers are shorter than the optimal AI-citation length:**
- Sitewide analysis of all rendered FAQ answers across `_docs/en/*.md` (809 visible answers): average 25.9 words, median 25 words. Only 49/809 (6%) fall in a reasonable 40-200 word citation band; **zero** fall in the 134-167 word "sweet spot."
- Frontmatter `faq:` answers (meant to be the longer, SEO-optimized version) average only 37 words, median 34 — barely longer than the visible answers, and only 7/815 land in the 100-200 word range, none in 134-167.
- One clear exception: the `tasklist-*` page series (e.g. `tasklist-01-add-tasks.md` / `/guide/add-task/`) has been recently reworked with frontmatter FAQ answers running 60-120+ words — these read as a deliberate "round" of GEO optimization (consistent with the `project_howto_schema` memory note about an in-progress AI-citation improvement effort). The other ~130 pages have not yet received this treatment.
- Shortest average-answer pages (candidates for enrichment first, since they're likely to be under-cited): `tasklistad-12-refer-tasks.md` (9.3 words avg), `search-01-blink-search.md` (18.3), `sidebar-03-recent.md` (19.8), `mobile-08-mobile-troubleshooting.md` (20.0), `account-04-email-notifications.md` (20.5).

Short, tight answers aren't wrong for UX (that's the intent of `GUIDE-FORMAT.md`), but AI engines cite longer self-contained passages (134-167 words) more often because they carry more standalone context. The frontmatter `faq:` field already exists as the correct place to carry this longer version without touching the concise visible copy — it just needs to actually be written to that length on the ~130 pages that haven't had the enrichment pass yet.

## Structural Readability

- H2/H3 conventions are documented and consistently applied (`## Add a Task`, `### In the Workspace` — topic labels, not questions), per `GUIDE-FORMAT.md`.
- Question-based headings are deliberately confined to the FAQ section only. This is a defensible house-style choice (keeps body nav scannable) and is compensated for by the `FAQPage` schema explicitly encoding the questions — but it does mean body H2/H3s alone won't match natural-language queries as directly as question-phrased headings would.
- Numbered steps used for sequential procedures (good for AI extraction of "how to" answers).
- `<br>` spacing rule after lists followed by paragraph text is a minor formatting nicety, not a GEO factor.

## Multi-Modal Content

- Every page checked includes descriptive screenshots with meaningful alt text (e.g., "Add sibling task in Quire Tree view") — good baseline image accessibility.
- No embedded video/YouTube content found within guide pages themselves, despite Potix having an active YouTube channel (`youtube.com/@QuireIoOfficial`) referenced in the organization's `sameAs` schema. YouTube mentions/embeds have the strongest observed correlation with AI citations (~0.737) of any brand signal — this is the single largest content-format gap in the audit.
- No downloadable/structured data assets (tables, comparison charts) observed beyond screenshots and text.

## Authority & Brand Signals

**What works well:**
- Every page carries a shared `Organization` JSON-LD (`@id: https://quire.io#organization`) with a substantial `sameAs` array: Facebook, Twitter/X, Instagram, LinkedIn, YouTube, G2, Capterra, GetApp, SoftwareAdvice, Trustpilot, and a **Traditional Chinese Wikipedia entry** (`zh.wikipedia.org/zh-tw/Quire`).
- `WebSite` schema names the guide clearly ("Quire & You") and every page's `og:site_name`/`title` consistently anchors "Quire" as the product name.
- Canonical tags + full 13-language hreflang alternate links present on every page — strong entity consistency across locales.
- Visible "Last updated:" date on-page plus `datePublished`/`dateModified` in schema.

**Gaps:**
- **No English Wikipedia entity.** `en.wikipedia.org/wiki/Quire` resolves to an unrelated disambiguation page (paper measurement unit, a historical person, a comic character, "Quill & Quire" magazine) with no mention of the Potix/Quire software. Since AI answer engines weight Wikipedia presence heavily as an authority signal, and English is the guide's primary/source language, this is a meaningful gap relative to the strong zh-TW Wikipedia entry already secured.
- Reddit presence could not be confirmed via unauthenticated API access in this audit (Reddit's search API requires auth/JS rendering) — recommend a manual or DataForSEO-assisted check, since Reddit is a high-correlation signal for AI citation and is unusually cheap to build (community engagement, not code changes).
- No visible named author/reviewer byline on pages (only a generic `article:author` pointing at a Facebook URL) — acceptable for a corporate help center, but adding a "Reviewed by the Quire Team" line with a review cadence would strengthen E-E-A-T signals cheaply.
- Domain Rating/backlink signal not evaluated in this pass (would require a backlink data source); noted as the weakest-correlation signal (~0.266) regardless, so it's not a priority.

## Technical Accessibility

- Confirmed fully server-side rendered: `curl` (no JS execution) on `/guide/add-task/` returns the complete rendered HTML including body content, FAQ text, and both JSON-LD blocks — no SPA shell, no client-side hydration dependency. This is the ideal state for AI crawlers, which generally do not execute JavaScript.
- `sitemap.xml` is a valid sitemap index correctly separating main site, blog, and guide sections.
- Canonical URLs and complete `hreflang` alternate sets present on every page checked.
- robots.txt imposes `Crawl-delay: 5` on ClaudeBot and `Crawl-delay: 10` on Bytespider — reasonable, won't meaningfully throttle indexing at this site's scale.
- No `llms.txt` (see above) — the only technical-accessibility item not fully closed out, and it's optional rather than blocking.

## Brand Mention Analysis Summary

| Signal | Status | Correlation weight |
|---|---|---|
| Wikipedia (zh-TW) | Present | High |
| Wikipedia (English) | **Missing** | High |
| YouTube channel | Present (linked in schema) but **not embedded in guide content** | ~0.737 (strongest) |
| Reddit | Unconfirmed (could not verify via unauthenticated tooling) | High |
| LinkedIn | Present | Moderate |
| G2 / Capterra / Trustpilot / GetApp / SoftwareAdvice | Present | Supports review-based trust signals |
| Domain Rating / backlinks | Not evaluated | Weak (~0.266) |

## Top 5 Highest-Impact Changes

1. **Lengthen frontmatter `faq:` answers toward the 100-167 word range on the ~130 pages not yet covered by the tasklist-series rework**, keeping the shorter visible FAQ copy untouched per `GUIDE-FORMAT.md`. This directly targets the single biggest citability gap (0% of current answers in the optimal range). *Effort: Medium — content work per page, can be templated/batched; no code changes.*
2. **Secure or correct an English-language Wikipedia entity for Quire (the software).** Currently `en.wikipedia.org/wiki/Quire` is an unrelated disambiguation page. Given the strength of the zh-TW article already in place, replicate that effort in English (subject to Wikipedia's notability/COI editing rules — likely needs a neutral third-party editor). *Effort: High — outside engineering, requires PR/legal-adjacent process, no direct code change.*
3. **Embed or link YouTube walkthrough videos directly within relevant guide pages** (not just in organization-level schema). YouTube mentions carry the strongest measured correlation with AI citation of any brand signal in this analysis. Start with the highest-traffic pages (add-task, board-view, permissions). *Effort: Medium — requires video assets (may already exist on the channel) plus a page-embed component.*
4. **Add optional `Disallow` rules for CCBot and anthropic-ai in robots.txt** if Potix wants to keep permitting AI-search crawlers (GPTBot, PerplexityBot, ClaudeBot) while opting out of training-only scraping. Currently all bots inherit the same wildcard `Allow: /`. *Effort: Low — a few lines in robots.txt.*
5. **Publish a `/guide/llms.txt`** listing canonical (English) URLs per topic with short summaries, deduplicating the 13-locale hreflang set into one clean ingestion index, and optionally asserting licensing terms via RSL 1.0. *Effort: Low-Medium — can be generated from existing frontmatter (`title`, `description`, `permalink`) via a build script.*

## Platform-Specific Estimated Scores (not live-tested)

| Platform | Estimated Score | Basis |
|---|---|---|
| Google AI Overviews | 80/100 | Strong FAQPage schema + SSR + sitemap; Google is generally most schema-literal |
| ChatGPT (GPTBot/OAI-SearchBot) | 70/100 | Crawlable and well-structured, but short FAQ answers limit long-form extractable citations |
| Perplexity | 75/100 | Crawlable (PerplexityBot inherits wildcard allow); benefits from clear dates and direct answers |
| Bing Copilot | 78/100 | Indexed via Bing, benefits from same schema/SSR strengths as Google |

These are directional estimates based on technical/content readiness only. Recommend running live checks (e.g., `ai_optimization_chat_gpt_scraper` / `ai_opt_llm_ment_search` via DataForSEO MCP, if enabled) to validate actual citation rates once available.