| name | bitw-seo-audit |
| description | Audit and fix SEO, meta-tag, and structured-data completeness for the behavior-in-the-wild.github.io academic website. Also the canonical reference for everything needed when adding a new page. Use when adding new pages, reviewing PRs, or running a full site health check. |
| metadata | {"author":"Yaman Kumar","version":"4.0.0","tags":["seo","structured-data","json-ld","academic","bitw"]} |
| compatibility | {"agents":["claude-code"],"requirements":["behavior-in-the-wild.github.io repo checked out locally","Python 3"]} |
BITW SEO Audit & New Page Guide
Canonical reference for the behavior-in-the-wild.github.io academic research site — both the automated audit tool and the human checklist for adding new pages.
Quick Start (audit)
python3 scripts/audit.py --repo ~/git/behavior-in-the-wild.github.io
python3 scripts/audit.py --repo ~/git/behavior-in-the-wild.github.io --fix
Site Structure
Repo: behavior-in-the-wild/behavior-in-the-wild.github.io (deploying org repo). PRs go here from the yamanksingla fork — never directly to org main.
File naming: preserve the original paper abbreviation casing (e.g. LCBM.html, SDR-Bench.html, memorability.html). No spaces; use hyphens for multi-word slugs.
Page types and their schema.org @type:
| Page | @type | og:type | Citation tags? |
|---|
index.html | WebSite | website | No |
| Paper pages (most) | ScholarlyArticle | article | Yes |
| Benchmark/dataset pages | Dataset | article | Yes |
| Dual paper+dataset | ["ScholarlyArticle","Dataset"] | article | Yes |
PersuasionArena.html | SoftwareApplication | website | No |
the-culture-repository.html | Dataset | website | No |
Redirect pages (cultural-alignment.html, transsuasion.html) | — | — | Skip entirely |
Publisher — per venue rule:
| Paper type | publisher value |
|---|
| Published at a known conference (ICLR, AAAI, WACV, CVPR, NeurIPS, EMNLP, EACL, ECCV) | Use the actual academic publisher for that venue (see CONF_MAP in audit.py) |
| Preprint with arXiv ID | {"@type": "Organization", "name": "arXiv", "url": "https://arxiv.org"} |
| Preprint/tool with no venue and no arXiv | {"@type": "Organization", "name": "Behavior In The Wild Research Group", "url": "https://behavior-in-the-wild.github.io"} |
Conference→publisher mapping:
Add new venues to CONF_MAP in both scripts/audit.py and scripts/fix_schema.py (or /tmp/fix_schema_full.py).
Canonical URL format: https://behavior-in-the-wild.github.io/{PageNameNoExtension} (case-preserved, no trailing slash). Exception: index.html → https://behavior-in-the-wild.github.io/.
Contact email: behavior-in-the-wild@googlegroups.com — include on every paper page.
Adding a New Page — Complete Checklist
1. File setup
- Copy the most structurally similar existing page as a template.
- Place at repo root:
{Slug}.html
- Place paper-specific images in
images/{slug}/ (lowercase slug for the directory).
- Use
images/Human-Behavior.png as og/twitter image fallback if no paper teaser exists.
2. <head> — required elements in order
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="description" content="...AI/LLM/behavior keywords...">
<meta name="keywords" content="...">
<title>Full Paper Title — Behavior in the Wild</title>
<link rel="canonical" href="https://behavior-in-the-wild.github.io/{Slug}">
<link rel="icon" href="https://cdn-icons-png.flaticon.com/512/954/954591.png">
<meta property="og:type" content="article">
<meta property="og:title" content="...">
<meta property="og:description" content="...">
<meta property="og:url" content="https://behavior-in-the-wild.github.io/{Slug}">
<meta property="og:image" content="https://behavior-in-the-wild.github.io/images/{slug}/teaser.png">
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:title" content="...">
<meta name="twitter:description" content="...">
<meta name="twitter:image" content="...same as og:image...">
<script type="application/ld+json">...</script>
<meta name="citation_title" content="Full Paper Title">
<meta name="citation_author" content="Last, First">
<meta name="citation_publication_date" content="YYYY">
<meta name="citation_conference_title" content="Full Conference Name">
<meta name="citation_arxiv_id" content="XXXX.XXXXX">
<meta name="citation_pdf_url" content="https://arxiv.org/pdf/XXXX.XXXXX">
</head>
SEO rule: every <title> and <meta name="description"> must contain AI/LLM/agentic qualifiers alongside "Behavior in the Wild" to prevent ambiguous n-gram matches (human behavior, in the wild) from attracting unrelated traffic.
3. Page body — section order
<nav class="site-nav">
<a href="./index.html">← Behavior in the Wild</a>
</nav>
<div class="page">
<h1>Full Paper Title</h1>
<div class="authors">
<a href="https://author-homepage.com/">Author Name<sup>*</sup></a>, ...
</div>
<div class="equal-contrib">* Equal Contribution</div> <!-- omit if not applicable -->
<div class="affiliation">
<img src="images/adobe-logo.png" alt="Adobe">
<a href="https://adobe.mdsr.live/" target="_blank">Media and Data Science Research (MDSR) Lab, Adobe</a>
</div>
<p class="venue">ICLR 2025</p> <!-- abbreviated venue + year, shown in red -->
<p>Get in touch at <a href="mailto:behavior-in-the-wild@googlegroups.com">behavior-in-the-wild@googlegroups.com</a></p>
<div class="paper-links">
<a href="https://arxiv.org/abs/...">Paper</a>
<a href="https://github.com/behavior-in-the-wild/...">Code</a>
<a href="https://huggingface.co/datasets/behavior-in-the-wild/...">Dataset</a> <!-- if applicable -->
<a href="https://...demo...">Demo</a> <!-- if applicable -->
</div>
<!-- ... content sections (Abstract, Method, Results, etc.) ... -->
<h2 id="BibTeX">BibTeX</h2>
<pre>@inproceedings{...}</pre>
<hr>
<footer>...</footer>
</div>
Author links: all authors must have clickable <a href> links to their personal/academic pages. Use Google Scholar or personal homepage — not LinkedIn where avoidable (LinkedIn URLs change; personal sites are stable).
Affiliation logos: the <div class="affiliation"> block may include logo <img> tags but they are optional and can be omitted for cleaner pages. The text link to the lab is required.
BibTeX: required on every paper page — Google Scholar scrapes it. Use <h2 id="BibTeX"> exactly so anchor links work.
4. JSON-LD block
Minimum complete example for a published ScholarlyArticle:
{
"@context": "https://schema.org",
"@type": "ScholarlyArticle",
"headline": "Full Paper Title",
"name": "Full Paper Title",
"url": "https://behavior-in-the-wild.github.io/{Slug}",
"image": "https://behavior-in-the-wild.github.io/images/{slug}/teaser.png",
"description": "One-sentence abstract with AI/LLM keywords.",
"datePublished": "2025",
"author": [
{ "@type": "Person", "name": "Author Name", "url": "https://author-homepage.com/" }
],
"keywords": ["keyword1", "keyword2"],
"isPartOf": {
"@type": "CreativeWork",
"name": "Proceedings of the International Conference on Learning Representations (ICLR) 2025"
},
"publication": {
"@type": "PublicationEvent",
"name": "International Conference on Learning Representations (ICLR) 2025",
"startDate": "2025"
},
"sameAs": "https://arxiv.org/abs/XXXX.XXXXX",
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@type": "Organization",
"name": "International Conference on Learning Representations",
"url": "https://iclr.cc"
}
}
Field rules:
headline = same value as name (required by Google for Article types)image = paper teaser URL; use og:image fallback if no teaserauthor[].url = personal/academic homepage per author (improves entity disambiguation in Scholar)isPartOf for published paper: {"@type": "CreativeWork", "name": "Proceedings of {Full Conference Name} {Year}"} — CreativeWork is correct for proceedings (one-time publication), not Periodical (which implies a recurring journal)isPartOf for preprint: absent — never invent a venueisPartOf for Dataset: {"@type": "DataCatalog", "name": "Behavior In The Wild", "url": "https://behavior-in-the-wild.github.io"}publication: PublicationEvent captures the actual conference event for semantic search; use the same full name as isPartOfsameAs: arXiv canonical URL — required when arXiv ID is known; signals authoritative source to Googlelicense: always CC BY 4.0publisher: per-venue (see Publisher rule above) — actual academic publisher for conference papers, arXiv for preprints with arXiv ID, BITW Research Group for everything else
For Dataset pages, replace author with creator. For dual-type ["ScholarlyArticle","Dataset"], include both author and creator (same people), and use DataCatalog for isPartOf (Google Dataset Search requires it).
5. Add to index.html contributions list
Entry format (newest papers go at the top):
<li>
<a href="./{Slug}.html">Full Paper Title</a>
<span class="venue-year">— {VenueAbbr} {Year}</span>
<span class="award">· {Award Name} 🏆</span>
<span class="tag tag-{type}">{Tag Label}</span>
</li>
Available behavior tags:
| CSS class | Label | Color | Use when... |
|---|
tag-explain | Explaining Behavior | blue (#5a7099) | Paper analyzes/models why humans behave a certain way |
tag-predict | Predicting Behavior | green (#5f8a62) | Paper predicts human responses, memorability, preferences |
tag-optimize | Interventions to Optimize Behavior | brown (#a8765f) | Paper generates/optimizes content to influence behavior |
tag-understand | Using Behavior To Understand Other Modalities Better | purple (#8a6fb0) | Behavior signal improves a non-behavior task (VLM, NLP, etc.) |
tag-learn | Learning from Human Digital Traces | olive (#7a6a3f) | Learning from implicit behavior signals (clicks, gaze, engagement) |
Venue abbreviation format: use the widely recognized short form — ICLR, AAAI, NeurIPS, WACV, EMNLP, EACL, ECCV, CVPR, etc. Preprints → "Preprint". PhD thesis → "PhD Thesis".
6. sitemap.xml
The automated sitemap bot (create-pull-request/sitemap branch) usually updates this on merge. If adding manually, add an entry:
<url>
<loc>https://behavior-in-the-wild.github.io/{Slug}</loc>
<lastmod>{YYYY-MM-DD}</lastmod>
</url>
Update <lastmod> to today's date for any page you touch.
7. PR
- Branch off
upstream/main; PR target is behavior-in-the-wild:main (the org repo) from the yamanksingla fork
- Run the audit script before opening:
python3 .claude/skills/bitw-seo-audit/scripts/audit.py --fix
- Review
git diff — verify isPartOf name matches the exact conference from the Highwire tag, sameAs URL is correct
Automated Audit — What Is Checked
Every content page
| # | Check | Rule |
|---|
| 1 | <html lang="en"> | Must be present |
| 2 | <title> | Present and non-empty |
| 3 | <meta name="description"> | Present; should contain AI/LLM keywords |
| 4 | <link rel="canonical"> | Must match https://behavior-in-the-wild.github.io/{Slug} |
| 5 | og:title, og:description, og:url, og:image | All four required |
| 6 | twitter:card, twitter:title, twitter:description | All three required |
| 7 | og:type | article for papers; website for home/tool/dataset-only pages |
| 8 | JSON-LD present | At least one <script type="application/ld+json"> block |
| 9 | JSON-LD license | CC BY 4.0 on every non-WebSite block |
| 10 | JSON-LD headline | Required on ScholarlyArticle (same value as name) |
| 11 | JSON-LD image | Required on ScholarlyArticle |
| 12 | JSON-LD isPartOf | CreativeWork for published paper, DataCatalog for Dataset, absent for preprints |
| 13 | JSON-LD publication | PublicationEvent required on published ScholarlyArticle |
| 14 | JSON-LD author | Required on ScholarlyArticle |
| 15 | JSON-LD creator | Required on Dataset |
| 16 | JSON-LD publisher | Per-venue: conference publisher, arXiv for arXiv preprints, BITW for others |
| 17 | JSON-LD sameAs | Required when citation_arxiv_id Highwire tag is present |
Paper/benchmark pages only
| # | Check | Rule |
|---|
| 15 | citation_title | Highwire tag required |
| 16 | citation_author | At least one Highwire author tag |
| 17 | citation_publication_date | Highwire date tag required |
What Is Auto-Fixed (--fix / fix_schema_full.py)
- Missing
license → inserts CC BY 4.0
isPartOf: Periodical/Website on ScholarlyArticle → replaced with CreativeWork (from Highwire venue tag) or removed (preprint)isPartOf on Dataset → replaced with DataCatalog- Missing
headline on ScholarlyArticle → set to name value
- Missing
image → set from og:image meta tag
- Missing
author[].url → matched from <div class="authors"> link tags in page body
- Missing
publication: PublicationEvent on published ScholarlyArticle → added
- Wrong
publisher → replaced with per-venue publisher (CONF_MAP lookup)
- Missing
sameAs when citation_arxiv_id present → adds arXiv URL
- Missing
creator on Dataset/dual-type → copied from author
Not auto-fixed (require per-page content knowledge):
- Missing/thin meta description, title, OG/Twitter tags, twitter:image
- Missing Highwire citation tags or
citation_pdf_url
- Author URL mismatches (name in JSON-LD differs from HTML body link text)
- Wrong
og:type, missing lang="en", missing BibTeX section
Trigger Phrases
Should trigger:
- "audit the BITW website"
- "check SEO on behavior-in-the-wild"
- "make sure all pages have uniform optimizations"
- "new page added to BITW, check it"
- "what do I need to do to add a new page to BITW?"
- "run the SEO health check on the site"
- "fix structured data issues on the academic site"
- "does this new page follow all the BITW best practices?"
Should NOT trigger:
- "review this PR for bugs"
- "update the homepage copy"
- "fix merge conflicts"
