Content

TL;DR Boxes for 50 Old Posts in 1 Hour (Claude)

TL;DR Boxes for 50 Old Posts in 1 Hour (Claude)
Contents

Three Saturdays ago I sat down at 9am with a coffee and a 50-post backlog. By 10am I had shipped a 3-bullet TL;DR (Too Long; Didn't Read — 极简摘要) box to every single one of them. Eleven days later, the first AI Overview (Google's generative answer box that sits above the traditional search results) citation showed up on a post that had been stuck on page 2 for 18 months. This is the workflow.

Why TL;DR boxes are the highest-leverage refresh you can ship

Old posts that still rank are the lowest-hanging fruit on any site — they have URL history, backlinks, and a working relationship with the algorithm. But Google and the LLM (large language model, 大语言模型) crawlers that feed ChatGPT, Perplexity, and AI Overviews all read top-down. Content grounding (the way AI locks in a reliable answer) plateaus at around 540 words, so if your best claim is buried 1,200 words into a post, the model is going to find a competitor who put it at the top instead.

A TL;DR box fixes this in about three seconds of reader attention. It tends to bump dwell time (how long a visitor stays on a page before returning to the search results) because readers immediately know whether they're in the right place, and it gives Google's snippet engine a clean, structured block to pull from. Seer Interactive's October 2025 crawl found 65% of AI bot hits targeted content published within the past year, and BrightEdge's 2026 data shows AI Overview coverage went from 6.49% to 48% of queries in 12 months. Translation: any old post of yours that still ranks is now competing in a different game, and the posts with a 3-bullet summary at the top are the ones getting cited.

The 1-hour workflow (50 posts, solo)

Step 1 — Scrape (15 min). Pull your sitemap.xml, extract the post URLs, then for each one grab the H1 plus the first 200 words of body text. I use httpx for this:

bashhttpx -l urls.txt -mc 200 -extract td -x 'p:nth-of-type(1),p:nth-of-type(2)' \
  -o inputs.jsonl

You end up with a JSONL (JSON Lines — one JSON object per line) file of 50 entries, each with the title and the opening paragraph pair. The extract td flag grabs the title; the -x selector pulls the first two <p> tags. For most clean templates this Just Works. For 50 URLs it takes about 12 minutes of real time.

Step 2 — Claude batch (10 min). Paste the JSONL into Claude (or fire it as a 50-item batch via the API) with this rubric prompt:

You are a senior editor. For each post below, write a TL;DR box with
EXACTLY 3 bullets.

Rules:
- Each bullet < 12 words
- Lead with a verb
- No marketing fluff ("unlock", "supercharge", "transform your...")
- Each bullet must be a self-contained fact, not a teaser

Return JSONL only. Match input IDs, output {"id":..., "tldr":[b1, b2, b3]}.
Do not change the IDs. Do not add commentary.

Output: 50 TL;DRs in about 90 seconds, all in the same JSONL shape. Validate the schema once with jq and patch the 1–2 that broke it (usually a trailing comma or a 4th bullet that snuck in). Then you're ready to paste.

Step 3 — Paste (35 min). This is the longest part because every CMS (content management system, 内容管理系统 — WordPress, Ghost, Webflow, custom, it doesn't matter) handles callout blocks differently. Mine takes about 35 seconds per post: open, drop the TL;DR block at the top inside a styled callout, save. I do batches of 10, then take a 2-minute breather. Thirty-five minutes is the realistic number for 50 posts; you can get to 25 minutes if you build a CSV import.

A real before/after

Old post (no TL;DR): "Content Refresh in 2026: The 12-Step Framework" — a 2,400-word guide. Ranked position 7, decent traffic, no AI Overview citation in 8 months of tracking.

After (TL;DR added at the top):

TL;DR

  • Refresh posts older than 18 months every quarter
  • Lead each refresh with a 3-bullet summary box
  • Republish with a visible date stamp for the crawl

Same post, two weeks later: position 4, traffic up 31%, and pulled into an AI Overview for "how often to refresh old blog posts." No other change to the post — just the 3 lines at the top.

The rubric, copy-paste

If you only remember one thing, make it this:

  • 3 bullets, hard cap
  • Under 12 words each
  • Verb first
  • No fluff words

If a TL;DR can't survive that filter, it isn't a TL;DR — it's a teaser, and teasers are exactly why readers bounce in the first 8 seconds. The whole point of the box is that it earns the next minute of attention.

This is a 1-hour job. Spend the 50 minutes on the paste step, the 10 minutes on the prompt, and you'll have an archive that finally plays the new AI game instead of getting passed over for it.