Content strategy for AI search: lessons from 290 articles

Content strategy for AI search: lessons from 290 articles

LA
Jedi Solomon
··4 min read

A content strategy for AI search is a large set of specific pages, each answering one buyer question in its first two sentences, with a person checking every page before it goes live. We ran that approach on a Shopify store and published about 290 articles. The results were uneven in useful ways.

What the 290 articles did for one store

JCS Fashions is LynkAI's first customer, a 4,878-product Shopify apparel store in Milpitas, California. From April to September we published about 290 quality-gated articles for it. Today 233 guide pages earn impressions on 24,164 queries.

Guide-page clicks went from 0 in February to 231 in August. Google clicks overall went from 205 a month in January to 626 in September. Sessions from AI assistants went from 20 in January to 569 in October.

Orders are a smaller story. AI-assistant orders were 5 in October. Shopify total sales rose from $1,706 in January to $26,371 in October, but that figure covers all channels, paid ads included, and October 2025 was $16,558, so the season explains part of it. Sales referred by search engines went from $283 to $3,212 over the same months. The best sales day on record was Oct 18, 2026, at $5,145 on 156 orders, against a previous record of $3,517 on Sep 7, 2025. Assistants send visits faster than they send orders: 569 sessions in October against 5 orders. The full numbers are on the JCS Fashions customer page.

Write one page per question

Divide 24,164 queries by 233 pages and you get roughly 100 queries per page. That only happens when a page answers one question precisely enough that many phrasings land on it. Sizing guides, fabric and care explainers, "what to wear to" guides, and shipping and returns explainers all fit. A page called "Our guide to summer" does not.

ChatGPT, Gemini and Perplexity lift passages. A passage that opens with the answer can be quoted whole. A passage that opens with a story about the brand has to be dug through before anything useful turns up, so a competitor's cleaner page is easier to quote. Open every page with the answer and put the story lower down, where a reader who wants it will find it.

Rank 11 to 100 still gets cited

Ahrefs (March 2026) found that 38% of AI Overview citations come from the top 10 results and 31% from positions 11 to 100. A new page on a small store spends its first months in that second group, and it can still be used. Ahrefs also measured, across 300,000 keywords, that the first Google result loses about 58% of its clicks when an AI Overview appears. Position one is worth less than it was, so a content strategy for AI search can reasonably aim at breadth across many questions.

What we stopped spending time on

llms.txt was the first thing to go. In May 2026, 97% of llms.txt files received zero AI crawler requests, and Google discourages the file.

Structured data was the second. It correlates only weakly with AI citations. Product schema still belongs on product pages for rich results and feed consistency, but we do not expect it to carry a guide page into an answer. The time is better spent on the sentences the engine will quote.

Review every page before it publishes

A person reads every page we publish. Ecommerce answers go wrong in specific ways: a return window that differs from the policy page, a fabric claim the product page contradicts, a size chart that disagrees with the guide. The reviewer needs the return policy, the size charts and the catalog open next to the draft. Once an assistant quotes the wrong number, the customer arrives expecting it. A wrong figure on a page that gets quoted does more harm than a thin page nobody quotes, so accuracy comes before volume.

A content strategy for AI search you can start this week

  1. Export 90 days of queries from Google Search Console. Filter to questions (how, what, which, can, does) that have impressions and no dedicated page.
  2. Pick five. Write each as answer-first content: the answer in the first two sentences, then the detail, then links to the matching collection or product pages.
  3. Check every fact against your catalog and your policy pages. Fix the page or the policy, whichever is wrong.
  4. Confirm that crawlers can read the pages. Look at robots.txt and make sure nothing blocks the guide directory.
  5. Write down 10 buyer questions and ask them in ChatGPT, Gemini and Perplexity. Repeat after four weeks and note which brands are named.

If you want to see where your store stands before you write anything, the free AI visibility scan shows whether those engines name you today.