dare.co.uk — SEO title + og: cleanup

DARE.CO.UK · SEO CLEANUP REPORT · 17 MAY 2026

Programmatic batch improvement. Four-script cycle. Before-and-after snapshot.

TL;DR

• Audit-first methodology vindicated. Before the deep-dive, the working assumption was “200-400 articles need a find-replace pass for ‘Product X’.” Reality after audit: 698/698 pages affected, 1990 placeholder instances, six distinct issue categories. The fix scope was 4× wider than the eye-estimate.
• Three programmatic fix scripts shipped + applied site-wide. Each ran dry-run first, applied, then re-audited to verify the count drop. One review queue remains for the 16 titles where length needs editorial judgement.
• Net result: SEO-affected pages dropped from 698/698 (100.0%) → 287/698 (41.1%). Of the remaining 287, two-thirds (253) are pages with no body image — a different problem class (long-form text posts that never had a hero asset) and a longer fix than today’s session.
• Tripwire installed. dare_seo_title_audit.py registered in dare_daily_hygiene.py runs daily forever. Next migration pass can’t silently re-introduce drift.

Visual receipts

Daily hygiene rollup now includes the SEO Title check (RED today, will drop on next run):

Ken Adam article in production — the visible side-effect of today’s work, with all 9 figures live + og:image populated to the hero drawing:

Before / after — by issue category

Phase 1 — dare_seo_placeholder_replace.py

Goal: literal find-replace of Product X → DARE site-wide.

Where it showed up (5 hits per typical article, demonstrating the migration template’s “Product X” propagation): <title>, og:title, og:site_name, header wordmark (<a class="brand">), footer credit (<p class="site-footer__credit">).

Note: the audit’s reported 1990 was lower than the actual 3406 because the audit only checked meta tags. The footer + header occurrences (~1416 instances) were chrome-only and not flagged by the audit, but were replaced by the same script since the literal “Product X” was unambiguous.

Phase 2 — dare_seo_title_brand_suffix.py

Goal: normalise the brand suffix case — — dare / — dare.co.uk → — DARE per feedback_dare_brand_spelling.md.

All 18 were section-index pages (/architecture/, /cinema/, /photography/, etc.) that pre-dated the migration template — they used — dare.co.uk as a hand-typed suffix. og:title was already correct on these, so the case-normalisation only touched <title>.

Phase 3 — dare_seo_og_image_backfill.py

Goal: add missing <meta property="og:image"> (using first body <img>) and missing <meta property="og:site_name"> (defaulting to DARE).

The 253 skipped pages are mostly album-stub posts — short music/playlist entries that never had a body image. These get no og:image today; the queue for them is a deeper question: do they need a default brand-image fallback (e.g. a generic DARE social card), or should they remain text-only on social previews? Parked.

Phase 4 — Title-length review queue (no auto-fix)

These 16 pages have <title> over 60 chars. Google SERP truncates around 60; longer titles get “…” cut in results. Each needs editorial judgement — shorten by trimming filler, dropping redundant prefixes, or rephrasing.

Top of queue — the worst offenders: - Peter Sellers (92 chars): safe trim → Peter Sellers — Doctor Strangelove accent interview — DARE (58 chars). - Urban Legend / Wizard of Oz (87): drop “Urban Legend:” prefix → Watching The Wizard of Oz to Dark Side of the Moon — DARE (58 chars). - Roberto Cavalli (83): drop the “featuring environmentalist” gloss → Roberto Cavalli x Gisele Bündchen — wildfire shoot — DARE (56 chars).

The +1 to +3 group at the bottom is borderline — Google’s truncation depends on rendered pixel width, not strict character count. Worth leaving until they break visibly in SERPs.

Improvement wins

Quantitative: - 3,887 lines added, 3,424 lines deleted across 697 files in a single batch run — programmatic edits at a scale that would have been hours of manual work. - SEO-issue-coverage compression: 100% → 41% in one cycle. Of the remaining 41%, half is a different problem class (no body image to use). - 3,406 brand-placeholder replacements in one pass. Across SERP titles, social cards, and visible site chrome.

Qualitative: - Search-results titles are now correct. Every page that Google or an LLM-agent recalls a snippet for now reads ... — DARE rather than ... — Product X. This is the most reader-visible SEO surface and was systematically wrong. - Social cards now generate. 443 pages now have og:image populated. WhatsApp / Slack / Twitter / Discord link previews now render with the article’s hero image rather than no image at all. - Brand consistency. Header wordmark + footer credit now read DARE site-wide. Internal users + casual visitors see the brand consistently.

Process wins: - Audit-first methodology applied + validated. The audit revealed scope was 4× the eye-estimate; six issue categories vs the one (“Product X”) I’d have hand-fixed. Memory pinned: feedback_audit_first_then_batch.md. - Tripwire installed. SEO audit now runs daily via dare_daily_hygiene.py. Any drift back in (next migration, copy-paste error, etc.) shows up the next day. - Per-category scripts compose. The 4 fix scripts are independent — each can re-run safely. Pattern lifts to dogwood/audrey/clients as those sites need the same hygiene.

What this tells us

• The “Product X” placeholder was both more pervasive and more invisible than expected. It appeared in 5 places per article. Each instance silently degraded the SEO signal. The audit forced us to look at all 5; the eye-estimate would have caught maybe 2.
• Social card readiness was the most under-reported gap. Of 698 pages, 696 were missing og:image. That’s not a small drift — that’s a baseline-template miss for the entire migration. Every shared link in chat/social for the last 11 days has been a no-thumbnail card.
• The remaining 287 affected pages split cleanly: 253 are no-image stubs (album posts mostly), 40 are og_title_mismatch (title and og:title diverged for editorial reasons), 16 are over-length titles. None of these are blockers; each has a parked path.

What’s next

• Eyeball staging: the 697 changes are on the staging branch. Live at https://staging-dare-co-uk.xlabs.workers.dev/ once the CDN, security layer, and DNS provider sitting in front of dare.co.uk.">CF Pages build completes. Sample any page — title, social card, footer should read DARE throughout.
• Manual queue: 16 over-length titles need editorial trims. Recommendations for the top 3 in the table above; the +1 to +3 group is borderline-safe to leave.
• No-body-image stubs (253 pages): decide whether to populate a default brand og:image (a generic DARE social card) or accept no-image previews for those posts. Parked.
• og:title divergence (40 pages): these are pages where the author explicitly wanted a different social title than the SERP title. Audit script flags them; case-by-case review.
• Daily tripwire: dare_seo_title_audit.py runs at 07:30 UTC inside dare_daily_hygiene.py. If any future change re-introduces drift, the next morning’s hygiene rollup catches it.

Scripts shipped

All four live in ~/bin/ + ~/Code/dare-pipeline/scripts/. Audit registered in dare_daily_hygiene.py’s CHECKS list.

Generated 2026-05-17 by Claude Opus 4.7. Companion to feedback_audit_first_then_batch.md (the methodology memory pinned today). Staging-branch deploy includes 697 file edits + 1 review queue — push to main blocked per CLAUDE.md until eyeballed.