A practical prompt and report format for judging a page against a query, spotting gaps fast, and turning intuition into specific fixes.
ChatGPT’s Agent mode is useful as an SEO tool because it can do the boring browsing for us, and then hand back something we can argue with. Agent mode helps because it can open the search results, sample the pages we are competing with, and compare them to ours in the same sitting. To that end, if you have it available, here is a copy-and-paste prompt to try.
[PROMPT START]
You are a Google Quality Rater and senior SEO auditor.
Query (search term): [PASTE QUERY]
Target market: [COUNTRY/REGION], language: [LANGUAGE]
Our page to evaluate: [PASTE URL]
Task:
1) Open a fresh browser session and run the query. Capture the top organic results (at least top 5). Note any SERP features that change intent (AI answers, local pack, shopping, video, etc.).
2) Summarise the dominant user intent in one paragraph. If there are multiple intents, name the primary and secondary.
3) Evaluate our page against that intent using a quality-rater mindset:
– Helpfulness and completeness for the query
– Evidence of experience, expertise, author credibility, and trust signals
– Transparency (who we are, how to contact, policies where relevant)
– Content quality (originality, clarity, accuracy, depth)
– Page experience (speed signals you can observe, intrusive UX, ads, layout, mobile friendliness)
– Spam or low-quality signals (thin content, template-y pages, over-optimisation)
4) Compare our page to the top 5 ranking pages. For each competitor, explain why it is winning, using specific page elements (headings, depth, tools, media, internal links, author info, trust cues, etc.).
5) Give a “why we lose” diagnosis in plain English.
6) Provide fixes in two sections:
A) Page-level fixes (content, UX, intent match) ranked by impact and effort.
B) Site-level fixes (authority, trust, coverage, internal linking) ranked by impact and time.
7) Deliver a final output in this structure:
– SERP intent summary
– Competitive pattern (what winners share)
– Our page: strengths, weaknesses
– Gap list (what we lack vs winners)
– Priority actions (top 5) with rationale
– Risks and unknowns (what you could not verify)
– Notes and evidence (quotes or observed details, short and specific)
Constraints:
– Do not invent facts. If you cannot confirm something, say so.
– Prefer observations over theory. Use concrete examples from the pages you viewed.
– Keep recommendations practical. Avoid vague advice like “improve E-E-A-T” without saying how.
[PROMPT END]
What should we expect back?
Firstly, a good assessment of the search results. If the top results are all ‘guides’ and our page is a news update, the agent should call that out. Likewise, if the query screams ‘freshness’ and the results page is full of recent dates, we should not be surprised when our old static page struggles.
Secondly, a competitive pattern. The useful observation here is what the winners have in common. The agent should describe those patterns, not just list URLs.
Thirdly, a blunt ‘why we lose’ paragraph. This is where agent mode should score. We want it to say things like: “Our page answers half the question, but competitors answer the follow-up questions people obviously have.”
Finally, an action list. Two groups matter most: the things we can change on the page this week, and the things that need broader site work. If everything is labelled “high priority”, the report has failed.
As ever, agent mode can still be confidently wrong, especially on things it cannot measure precisely (speed, indexing, links). So treat the output as an evidence-led briefing, then verify the key claims.