The new AEO writer will be measured on "citation share and answer accuracy for target prompts". Endor's largest public content set works against that: on 15 package version pages sampled across the sitemap, every one shows "Overall 0 /10" next to healthy sub-scores (j2objc-annotations reads Security 10, Activity 8, Popularity 10, Quality 5, Overall 0). 13 of the 15 titles drop the version number ("mkdirp | Endor Labs") and the URLs are database IDs. An assistant asked "is mkdirp safe" can cite Endor saying zero out of ten, and the pages compete with each other for the same package name.
First action
Open ten version pages and log overall vs sub-scores to confirm the pattern.
How to measure it
Correct, versioned package pages
- Now
- –
- Target
- 100%
- By
- 30 days
Steps
Confirm the pattern
Sample 50 version pages and log scores and titles.
Technical Writer & Editor (AEO) 1 day
Find the cause
Compare the /version and /ai-model templates and data sync.
Digital (web) team 2 days
Fix the score
Render the real overall score or hide it when missing.
Digital (web) team 2 days
Rewrite titles
Ecosystem, package, version in title and H1.
Technical Writer & Editor (AEO) 2 days
Readable URLs
Slugs with 301s from ID URLs.
Digital (web) team 3 days
Canonicals
Older versions point to the latest version page.
Digital (web) team 1 day
Answer-first summary
Score, CVEs, reachable-risk note, AURI link.
Technical Writer & Editor (AEO) 1 week
Re-test prompts
Ask assistants about 20 packages and record what they cite.
Technical Writer & Editor (AEO) 2 weeks
Why it works
Have engineering find why the overall score renders as 0 on /version pages (AI model pages show a real score, so the template or field mapping is the likely cause) and fix it in the CMS sync. Put package, ecosystem and version in every title, H1 and URL slug ("npm mkdirp 0.5.6 security score"), redirect the ID URLs, add a canonical from each version to the package's latest version, and add a short answer-first summary: overall score, known CVEs, reachable-risk note and a link to scan with AURI. Re-check the target prompts afterwards.
These pages are Endor's widest footprint in search and AI answers. Correct scores turn them from a liability into proof of the platform's data, and clean titles and canonicals stop thousands of pages splitting the same query.
Ready to use
ChecklistPackage page repair list
[ ] 50-page sample logged (overall vs sub-scores, titles) [ ] Cause found in template or data sync [ ] Overall score renders correctly (or hidden when missing) [ ] Title/H1: ecosystem + package + version [ ] Readable slugs with 301 from ID URLs [ ] Canonical rules for older versions [ ] Answer-first summary block with AURI CTA [ ] 20 target prompts re-checked
Risks and how to handle them
Sync overwrites edits
Make the fix in the data pipeline, not by hand in the CMS.
Canonicals hide useful version pages
Keep vulnerable versions self-canonical where they carry unique CVEs.
Evidence
- Your websitePackage version pages, 15 sampled across the sitemap: every one shows "Overall 0 /10" beside healthy sub-scores (com.google.j2objc:j2objc-annotations: "Security 10 Activity 8 Popularity 10 Quality 5"; mkdirp: "Security 10 ... Popularity 10 Quality 4"). 13 of the 15 titles have no version number ("mkdirp | Endor Labs"), and the URLs are database IDs (/version/634a0bde982579fc3094a2c0). AI model pages show a real overall score (01-ai/Yi-1.5-34B: "Overall 6 /10").Source
- HiringTechnical Writer & Editor (AEO), Bengaluru, posted November 3, 2026: will "Execute Endor Labs' Answer Engine Optimization (AEO) strategy", create "how-to guides, integration walkthroughs, technical explainers, FAQs, glossaries, and comparison pages", "Refresh and optimize existing blogs, documentation, and website content", apply "answer-first writing, question-led headings ... schema markup", and track "AI visibility metrics, including citation share and answer accuracy for target prompts".Source
- Your websiteSitemap (November 4, 2026): 9,963 URLs, of which 3,040 are CVE pages (/vulnerability), 3,000 AI model pages (/ai-model), 2,994 open source package version pages (/version), 548 /learn articles, 180 event pages, 53 private event pages, 27 campaign landing pages (/lp), 22 use cases, 17 integrations and 2 /compare pages.Source
- Your websiteHomepage: "Endor Labs is a Leader in the Forrester Wave™: Agentic Development Security Tools, Q4 2026"; AURI for developers is "free forever" with install for Cursor, Claude Code, Codex, VS Code / Copilot and Antigravity; AI SAST claims to "cut up to 95% of false positives"; AURI learns from "AGENTS.md, CLAUDE.md" context files.Source