Verified against ChatGPT · 2026-08-09
Classify a keyword list by real search intent using what's actually ranking, not what the keyword sounds like it means
Reads the visible SERP pattern for each keyword — content type, format, and angle of the ranking pages — and classifies intent from that evidence, catching the keywords that sound informational but are actually served by product pages, or vice versa.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Classify search intent for the keyword list below using the evidence I give you about what's actually ranking, not just what the keyword phrase sounds like it means. KEYWORD LIST best running shoes for flat feet, how does plantar fasciitis form, running shoe size chart, Brooks Adrenaline vs Ghost SERP OBSERVATIONS (what's ranking for each, if you have it) "best running shoes for flat feet" — top 5 results are all e-commerce category/buying-guide hybrids with product grids, no pure blog posts in top 10. OUR PAGE TYPE OPTIONS Blog long-form articles, product category pages with filterable grids, comparison tables, a size-chart tool. RULE A keyword's surface wording is not reliable evidence of intent on its own — "best running shoes for flat feet" looks informational but if the visible SERP is dominated by e-commerce category pages and buying guides with embedded product grids, the real intent is commercial-investigation-with-purchase-readiness, not pure information-seeking, and a blog post competing against that SERP will underperform a well-built category page regardless of writing quality. If I've given you SERP observations for a keyword, base the classification on that evidence and say so. If I have not given you observations for a keyword, classify it from the keyword's wording as a provisional best guess, but label it explicitly as unverified and recommend checking the live SERP before committing a page type to it — do not present a wording-based guess with the same confidence as an evidence-based classification. FOR EACH KEYWORD, DETERMINE 1. Primary intent: informational, commercial investigation, transactional, or navigational. 2. Dominant ranking format if observed: long-form guide, comparison table, product/category page, tool/calculator, forum thread, video. 3. Which of our available page types actually matches that format, or whether none of them do. 4. Confidence: evidence-based or wording-based-guess. WHAT NOT TO DO Do not default every keyword containing "how," "what," or "best" to informational intent purely from the question word — question words are a weak signal on their own; the ranking format is the strong signal. Do not force a page type match if none of our available formats fits the observed SERP — say plainly that this keyword needs a format we don't currently have, rather than assigning it to the closest existing page type as if that were a fine substitute. OUTPUT FORMAT A table: Keyword | Primary intent | Dominant format | Matching page type (or "gap — need [format]") | Confidence. End with a short list of any keywords flagged as format gaps, since those represent either a page type we should build or a keyword we should deprioritize.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Left to its own defaults, GPT-5.1 classifies search intent almost entirely from the lexical pattern of the keyword phrase — question words toward informational, "buy"/"price" toward transactional — because that's the only signal available when no SERP evidence is supplied, and it will present that guess with the same confident tone as an evidence-grounded classification unless explicitly told to separate the two. That's the exact failure this prompt structure prevents: forcing a confidence label per row (evidence-based versus wording-based-guess) stops a plausible-sounding but ungrounded classification from being treated as equally reliable, which matters in SEO specifically because building the wrong page format for a keyword's real intent is a multi-week content investment that fails silently — the page can be well-written and still underperform because the SERP was never going to reward that format regardless of quality. Naming the dominant ranking format as a required output field, separate from the abstract intent label, is what actually operationalizes the classification into a page-type decision; "commercial investigation" alone doesn't tell a content team whether to build a comparison table or a long buying guide, but "top 5 results are category-page-plus-product-grid hybrids" does. The explicit refusal to force-fit a format gap into the closest available page type matters because an LLM asked to pick from a fixed list will pick the closest option even when none of them fit well, quietly hiding a real gap — that this site has no page type that can compete for this keyword at all — inside a recommendation that looks like a normal match.
What you get back
best running shoes for flat feet | Commercial investigation | Category page + buying guide hybrid | Product category page (existing) | Evidence-based. how does plantar fasciitis form | Informational | Long-form medical/health guide | Gap — need a dedicated editorial guide format | Evidence-based.
Verified against
ChatGPT GPT-5.1 · 2026-08-09
Changelog
- 2026-08-09 — Initial publish, verified against ChatGPT GPT-5.1.
Building this for real?
This is a free starting point. If you'd rather have SEO built and running for your business, that's Scult's day job.
EXPLORE SEO
