Day ???/100. Thursday. Coffee first.
Hey—It's Tim.
No story today. Here's the thing I said I'd give you.
That's the list I run when someone asks me why their content isn't landing. It's pulled out of a bigger catalogue of 600 that I've been building since March, and these are the hundred that matter on basically every site.
Go and open it. I'll explain what you're looking at.
What's actually in there.
Every row has the check, what it means in English, and how you know it passes. Not "improve your site speed." Actual criteria you can hold a page against.
Example, one of the two highest-priority AI checks in the whole thing:
AIO-025 — robots.txt does not block major AI crawlers
Passes when: robots.txt contains no Disallow rule matching '/' or any essential content path for GPTBot, ChatGPT-User, PerplexityBot, ClaudeBot, Google-Extended.
Go and open yoursite.com/robots.txt right now. Genuinely, right now, in another tab.
A lot of you are going to find a blanket disallow that someone added in 2023 when blocking AI scrapers was the fashionable thing to do. Nobody's looked at it since. You've been invisible to AI search for two years and it's been sat in a text file the whole time.
That's one row out of a hundred.
The breakdown, if you like numbers:
→ 74 are priority 1. Fix these before you touch anything else.
→ 56 are page-level. They apply to every single page you have.
→ 43 are sitewide.
→ 76 can be checked by a machine. The other 24 need a human looking at it.
Five recipes.
A hundred rows with no order is a to-do list you'll never start. So I've grouped them by the thing you're actually trying to fix. Filter the Recipe column.
1. "I want AI search to cite me" — 12 checks
Renders without JavaScript. robots.txt isn't blocking the crawlers. JSON-LD actually parses. Semantic HTML. Schema that resolves to a real entity. This is the one to start with if Monday's email annoyed you.
2. "Make one article extractable" — 8 checks
Everything from yesterday, as a list. Chunk size, standalone paragraphs, question headings with 40-50 word answers, BLUF. Run it on one article this morning and you'll see the pattern in your own writing immediately.
3. "Google isn't crawling my important pages" — 17 checks
Canonical chains and loops. Orphan pages. Infinite crawl spaces. Internal links routed through redirects. This is where most sites are quietly haemorrhaging crawl budget on URLs that shouldn't exist.
4. "My Core Web Vitals are red" — 20 checks
LCP in the initial payload, hydration bottlenecks, long tasks blocking the main thread, lazy-loading above the fold. The specific mechanical causes, not "make it faster."
5. "My traffic dropped and I don't know why" — 20 checks
The diagnostic sequence. Check you're not deindexed, check for manual actions, overlay algorithm dates on your traffic, separate branded from non-branded, look for 5xx errors served only to Googlebot. In that order, because the order saves you a fortnight.
The honest bit.
Doing this on one page is a nice morning's work. You'll learn something about your own site and you'll fix a few real problems.
Doing it across two hundred pages is a special kind of hell.
If you have 1000 or more pages, then I’ll be at church on Sunday, praying for you.
56 of these hundred are page-level. That means the check doesn't have an answer for your site, it has an answer for every page. "Do paragraphs stand alone without ambiguous references" isn't a question you answer once. It's a question you answer two hundred times, and by page forty you're not reading properly anymore.
That's not a discipline problem. It's the reason nobody does it, and it's the reason it's still worth doing.
I'll come back to that on Monday.
Send me your URL.
Reply with your homepage. Nothing else, just the URL.
I'll run AIO-025 against it myself and reply telling you what, if anything, is blocking you from AI search.
Takes me 30 seconds a site. No form, no signup.
Tomorrow I'm running this list against penfriend.ai and publishing whatever it finds. My own site, my own checklist, no editing out the embarrassing ones.
I'd rather show you that than a tidy case study about someone else.
One more time, in case you scrolled: the 100 checks are here.
See you tomorrow.
✌️ Tim "about to audit myself in public" Hanson
CMO @Penfriend.ai
Same brain, different platforms: X, Threads, LinkedIn.
P.S. If you open your robots.txt and panic, here's the distinction most people miss. GPTBot and CCBot are training scrapers, hoovering you up to build the next model. ChatGPT-User and OAI-SearchBot are live retrieval, which is the one that puts you inside an answer someone is reading right now. Blocking the first pair is a business decision you're allowed to make. Blocking the second pair is just making yourself invisible.
Penfriend.ai
Made by content marketers. Used by better ones.
