Claude can now
run your SEO

claude mcp add --transport http chatseo https://api.chatseo.app/mcp

Use the ChatSEO connector. If no site is set, list my sites
and ask me which one — do not guess.

Pull the last 28 days of Search Console data and build three lists:
- STRIKE ZONE — queries in average position 8-15 with 200+ impressions
- CTR GAP — pages with 500+ impressions whose CTR is under half of
  what their average position normally earns
- LOSING GROUND — queries whose average position dropped 2+ places
  versus the previous 28 days

Cap each list at 5 rows. Exclude brand queries everywhere — they move
for reasons that have nothing to do with the page.
For each row: query or URL | position | impressions | clicks | the
number that put it on this list.
If a figure is not in the data, write "cannot compute from this data".
Do not estimate it.

End with ONE opportunity: the page to work on, which list picked it,
and the single change to make.
Do not give me three options. Do not append a backlog.

Use the ChatSEO connector on my active site.

Ask me for the cutoff date if I didn't give one. Do not date a Google
update from your own knowledge — ask.

Compare the 28 days after the cutoff to the 28 days before, page by page.
Rank by absolute clicks lost, never by percentage — a page falling from
8 clicks to 2 is not a priority.
Top 5 losers only: URL | clicks before | clicks after | clicks lost | cause.

Name the cause from the data, and be specific:
- position dropped -> a content or authority problem
- position stable, CTR dropped -> a SERP problem: someone else's title,
  or a new AI overview sitting above you
- both stable, clicks still down -> impressions fell. That's demand,
  not ranking. Say so and stop.

Do not speculate beyond what the numbers support.
End with the ONE page to fix first, and why that one.

Use the ChatSEO connector. I'll give you a keyword or a competitor domain.

For a KEYWORD:
1. Pull search volume and the related queries that actually carry volume.
2. Pull the live SERP. Top 10: URL | page type | roughly how long |
   what the page is actually about.
3. Check my own Search Console — do I already rank for this, or anything
   adjacent? If yes, name the page and its position.
4. Say what the top 3 have in common that the rest don't. That is an
   observable pattern, not a theory about the algorithm.

For a COMPETITOR:
1. Pull their ranked keywords.
2. Cross them against my Search Console queries.
3. Report only the gaps — what they rank for and I don't — with their
   position and the page doing it.

Cap every table at 10 rows. Never invent volume, difficulty or traffic
figures: if the data isn't there, write "not available" and move on.

End with ONE angle: the page to write or update, the exact query it
targets, and the one thing it has to do better than the current #1.

Use the ChatSEO connector on my active site. I'll give you a URL.

1. Fetch the page and read its title and meta description as they are
   live right now. Do not work from memory.
2. Pull its top 10 Search Console queries by impressions, with position
   and CTR for each.
3. Tell me which query the page is actually being found for. It is often
   not the one it was written for.

Then write ONE title (max 60 characters) and ONE meta description
(max 155 characters), built from the real query rather than from the
page's own headline.

Hard rule: the title has to be true to what is on the page. If the page
doesn't deliver what the winning query wants, say so and stop — a better
title on the wrong page buys a bounce, not a ranking.

Give me one version, not three.
Close with the page's current CTR and the CTR its average position
normally earns, so I know what "it worked" will look like.

Use the ChatSEO connector.

Before touching anything:
1. Confirm the CMS is connected. WordPress: run the connection test.
   Shopify: list the blogs. If either fails, stop and tell me exactly
   what is missing — never attempt a partial publish.
2. Show me the before and after, field by field, and wait for my
   explicit "go". Never publish on an implied approval.

Then apply the change. Publish as a DRAFT unless I said "publish live".

Then VERIFY. This step is not optional:
- Re-fetch the live URL and read back the field you changed.
- Quote what is live now. If it doesn't match what you sent, write
  "the change did not land", show both values, and stop.

Finally, create a ChatSEO todo with: the URL, today's date, the field
changed, and the metric to check. That is what makes the result
measurable in 28 days.

End with one line: what changed, where, and when to check it.

Use the ChatSEO connector on my active site.

Compare the last 7 days of Search Console data to the 7 days before.
Under 250 words:
- total clicks and impressions, with the change
- the 3 biggest gaining pages
- the 3 biggest losing pages

Ignore any page moving fewer than 10 clicks — weekly noise is not news.

Then check my open ChatSEO todos. For any change shipped more than
28 days ago, say whether that page's clicks went up, down, or nowhere
since. If the window is too short to judge, write "too early" — do not
call it a win.

If nothing meaningful moved, write "quiet week" and stop. Do not
manufacture an insight to fill the space.
End with ONE action for this week.

When I ask a question about my SEO, answer it from the ChatSEO
connector, in plain English, in under 150 words.

Rules for every answer:
- Pull the data first. Never answer a question about MY site from
  memory or from general SEO knowledge.
- Always give the number and the period it covers. "Clicks are up"
  without a window is not an answer.
- If the data can't answer the question, write "cannot compute from
  this data" and say what would be needed. Do not estimate, do not
  extrapolate, do not round a guess into a figure.
- Correlation is not cause. If a page moved and I shipped a change,
  report both. Don't tell me the change caused it unless nothing
  else moved.
- No recommendation unless I asked for one. A question is a question.