Original insights by Chris Long
30-second rundown
Key learnings:
- AI search needs an accessible page before it needs a clever one: allow the crawlers you want, render important content in server-side HTML, and make the title, H1, URL, and page purpose agree.
- Every important section must survive alone: lead with the answer, use orderly headings, lists, tables, and FAQs, and write passages that remain clear when an AI system lifts them away from the page.
- Original evidence gives the page something worth citing: pair specific figures and first-hand findings with a named author, current facts, visible update dates, and coverage of the related questions buyers actually ask.
This week: Choose one commercial page and use the Chrome Web Developer extension to block JavaScript, then check whether the page’s key content still loads. Confirm the page title, URL, and H1 all communicate the same topic, and that the page’s main purpose remains clear without JavaScript. Then rewrite the opening answer and one important section so each makes sense on its own. Add a useful table, list, or FAQ only where it makes the information easier to compare or quote.
Coming 3 months: Use a crawler and AI Search monitoring workflow to check priority pages monthly for blocked bots, missing server-rendered content, stale facts, broken heading order, and changing fan-out questions. Assign one owner to review the alerts, make genuine updates, and record whether the corrected passages begin appearing in retrieval and citations.
Chris Long has rebuilt the old “perfectly optimized page” checklist for a world where the first reader may be an AI system deciding what to extract. His infographic turns the page into a border checkpoint with guards as the citation truck idles in the heat.
The caveat sits at the bottom: there is no perfect page or single ranking formula. This is a map of where effort may pay, not a holy tablet smuggled out of a server room.
1. Access is the first locked gate
A page cannot earn a citation from a crawler that cannot reach its useful content. Long recommends allowing relevant bots in robots.txt when that matches your policy, and placing essential copy in the page’s initial HTML because some AI crawlers may not execute JavaScript reliably.
Crawl access is the admission ticket. The title, H1, readable URL, and page subject should then point at the same intent. One page about running-shoe sizing should not wander into a half-drunk sermon about marathon nutrition, returns, and corporate history.
The sample page makes this alignment explicit: its title promises email-marketing automation, the H1 repeats that job, the URL stays readable, and the description previews pricing and evidence. None of this guarantees a citation, but confusion at the gate is a fine way to die before inspection.
2. Build passages that can leave home
Long places a direct four-to-six-sentence answer near the top because AI systems often extract passages rather than admire the whole page. He cites Kevin Indig’s analysis for a top-of-page attention pattern. Treat that as an observation, not a law carved into the desert floor.
Each important passage must make sense alone. Lead a section with its answer, then explain it. Use one H1, logical H2s and H3s, bullets, numbered steps, and tables when those formats genuinely clarify relationships.
Long’s earlier SaaS study found tables on 30% of sampled ChatGPT-cited pages versus 13% of roughly 25,000 Google-result pages, making tables 2.3 times more common in that comparison. Correlation is not causation, but structured information is easier to inspect than a swamp of uninterrupted prose.
Use real questions from sales calls and support tickets in FAQs, then answer immediately. Accurate machine-readable schema can describe the page, but decorative markup cannot rescue a vague answer wearing expensive shoes.
3. Give the system evidence nobody else owns
The web already has ten thousand definitions wearing identical grey suits. Long argues that a page earns distinction through attributable numbers, original research, audits, first-hand findings, and counter-intuitive results. Specific evidence creates a reason to cite you.
The fictional page shows pricing by contact tier, a 3.1-times result tied to a named benchmark, and migration findings competitors do not possess. A running-shoe retailer could publish wear-test results by surface and mileage instead of repeating that cushioning is “responsive,” a word so exhausted it should be found face-down behind the casino.
That evidence needs an identifiable author, consistent brand details, and a credible bio. It also needs to remain true. Use a visible last-updated date backed by dateModified markup, date-stamped figures, and facts that were actually reviewed.
Cover the related subquestions an answer engine may generate through query fan-out, but do not stuff every possible query into one bloated corpse. Keep this page focused and create other pages where complete coverage requires them.
The winning page is not “perfect.” It is accessible, coherent, extractable, specific, attributable, and alive enough to tell the truth today. Build that, and the citation inspectors may still wave another car through, but at least yours reaches the checkpoint with evidence in the glove box instead of sand.
