AI Prompt to Audit Blog Content for SEO via HubSpot's MCP Server
This HubSpot MCP prompt is for content marketers and SEO leads who want ChatGPT or Claude to pull live blog post data straight out of HubSpot and return a structured SEO audit — without exporting a CSV or copy-pasting post content by hand first. Once HubSpot's MCP server is connected to your AI client, the model can read a post's title, meta description, slug, word count, and topic cluster tags directly, then reason about what's actually hurting its search performance.
The prompt asks the model to check for the specific things that quietly tank organic traffic: a meta description that's missing or the wrong length, a title that doesn't match search intent, thin sections under target headers, and blog posts that have no internal links pointing to or from related cluster content. Because the data comes from the live HubSpot record rather than a stale export, the audit reflects what's actually published right now, which matters most for teams that update posts frequently.
This works well as a recurring check before a content calendar refresh, and it pairs naturally with tightening the audit instructions themselves — if the output feels too generic or keeps missing obvious gaps, running the prompt through Prompt Optimizer in Content mode is a reasonable next step to sharpen the constraints.
Prompt template
Want it sharper? Optimize this prompt with Prompt Optimizer, check it with the Prompt Debugger or shorten it with the Token Optimizer.
Example input
When to use it
- Running a quarterly SEO health check across an existing HubSpot blog
- Triaging which old posts to update before a content refresh sprint
- Reviewing a freshly drafted post in HubSpot before it goes live
- Spotting internal linking gaps between posts in the same topic cluster
Best practices
- Give the model the exact HubSpot MCP fields you want it to read (title, meta description, slug, body, tags) rather than letting it guess at post structure
- Ask for a prioritized list of fixes, not just a list of problems, so the output is actionable
- Cap the audit to one post per run at first so you can sanity-check the model's judgment before batching
- Have the model flag when it lacks enough data from HubSpot to make a call, instead of inventing missing metrics
Common mistakes
- Asking for a full-site audit in one prompt instead of auditing posts individually, which produces vague, generic feedback
- Not specifying the target keyword, so the model guesses search intent instead of checking against the actual one
- Treating word count alone as a quality signal instead of checking whether sections actually answer the query
- Forgetting to ask for internal link suggestions, which is one of the easiest wins an MCP-connected audit can surface
FAQs
Can ChatGPT or Claude actually read my HubSpot blog posts directly?
Only if your AI client is connected to HubSpot's MCP server, which lets the model call HubSpot's API to read post fields like title, body, and meta description. Without that connection, you'd need to paste the content in manually, which this prompt also supports.
What's the difference between this and just asking an SEO checker tool?
A dedicated SEO tool scores against fixed rules and competitor data. This prompt uses the model's reasoning against your actual target keyword and cluster structure, which is better for judgment calls like "is this section thin" but won't replace rank-tracking data.
How often should I re-run an SEO audit on the same post?
Most teams see value re-auditing a post every 3-6 months or after a meaningful traffic drop, since search intent and competing content both shift over time.
Which Cuelara tool can help me tighten this audit prompt if the output feels too generic?
Prompt Optimizer — its Content mode is built for sharpening vague content-review instructions into the kind of specific, structured criteria this audit prompt depends on.