Prompts to refresh content in Cursor
This checklist is based on How DigitalOcean updates its massive library of content 3-4x faster (without sacrificing quality) featuring Anish Singh Walia, published on the AI Lab by ActiveCampaign.
Get the prompt template
How to use these prompts
These are ready-to-use prompts adapted from the workflow Anish Singh Walia uses at DigitalOcean to refresh technical tutorials. Copy, paste, and swap in details where you see [BRACKETS]. They assume you’re working inside Cursor with the DataForSEO Model Context Protocol (MCP) server connected and your brief, style guide, and stale article open in the workspace. Treat the first output as a starting point and refine from there.
Prompt 1: Refresh the meta title, description, and opening paragraph
Best for: Starting a content refresh with the highest-impact on-page elements before touching body copy.
Use with: Cursor (with DataForSEO MCP connected)
Prompt 1
You are refreshing the article in @[ARTICLE_FILE].
Strictly follow the rules in @[SEO_BRIEF] and @[STYLE_GUIDE].
Using the DataForSEO MCP, pull live keyword stats, search volume, and SERP data for [PRIMARY_KEYWORD] in [LOCATION, e.g., United States]. Look at the top 5 organic results and identify what intent the SERP is rewarding right now.
Then give me:
1. An updated meta title (under 60 characters)
2. An updated meta description (under 155 characters)
3. A rewritten opening paragraph that matches current search intent
Apply the changes directly to @[ARTICLE_FILE]. Do not touch any other sections yet.
Variables to fill in:
- [ARTICLE_FILE] — The filename of the stale article in your Cursor workspace, e.g., kubernetes-monitoring-guide.md
- [SEO_BRIEF] — Filename of the brief with target keywords and ranking goals
- [STYLE_GUIDE] — Filename of your editorial guidelines
- [PRIMARY_KEYWORD] — The keyword the article targets, e.g., kubernetes monitoring
- [LOCATION] — Where you want SERP data pulled from. Anish defaults to United States, sometimes worldwide.
What to expect: A title and description that read naturally, mention the primary keyword, and reflect what the top results are doing. The opening paragraph should hook on a current search angle rather than a generic intro. If the output feels robotic, that is a signal that your style guide needs to be more specific.
Follow-up prompt:
The title and description are good but the opening paragraph still sounds generic. Rewrite it to lead with the specific problem [TARGET_AUDIENCE, e.g., DevOps engineers] are searching to solve in [CURRENT_YEAR/CONTEXT], based on the SERP data you pulled. No throat-clearing intros.
Prompt 2: Update one body section against live SERP data
Best for: Refreshing supporting sections one at a time, the way Anish works through an article.
Use with: Cursor (with DataForSEO MCP connected)
Refresh the section titled “[SECTION_HEADING]” in @[ARTICLE_FILE].
Follow @[SEO_BRIEF] and @[STYLE_GUIDE].
Using the DataForSEO MCP, pull the current SERP for [SECTION_KEYWORD_OR_QUERY] and identify:
- What subtopics the top 3 results cover that this section is missing
- Any outdated claims, versions, or numbers in the current section
- Questions related to this section that show up in People Also Ask
Update only this section. Keep the rest of the article untouched. Preserve the technical accuracy of any code, commands, or version-specific details. Flag anything you are not sure about with [VERIFY] so I can route it to a subject matter expert.
Variables to fill in:
- [SECTION_HEADING] — The exact heading text of the section you want updated
- [SECTION_KEYWORD_OR_QUERY] — The keyword or question this section answers
- [ARTICLE_FILE], [SEO_BRIEF], [STYLE_GUIDE] — Same files referenced in Prompt 1
What to expect: A revised section that closes specific gaps you can trace back to the SERP. If the output rewrites the whole article instead of one section, narrow your prompt and re-run. If sections lack depth, add more structured briefing documents rather than asking for “more detail.”
Follow-up prompt
You drifted into rewriting adjacent sections. Revert anything outside “[SECTION_HEADING]” to the original text. Then tighten that section to match the tone in @[STYLE_GUIDE] — shorter sentences, no inflated adjectives, and remove anything that reads as AI-generated filler.
Prompt 3: Generate an FAQ section from real user questions
Best for: Adding an FAQ that reflects how people actually search, not how an AI tool suggests expanding content.
Use with: Cursor (with DataForSEO MCP connected), paired with AlsoAsked or Answer the Public
Prompt 3
I’m adding an FAQ to @[ARTICLE_FILE].
Here are the real questions people ask about [TOPIC], pulled from AlsoAsked and Answer the Public:
[PASTE_QUESTIONS_HERE]
Using the DataForSEO MCP, pull SERP and People Also Ask data for [PRIMARY_KEYWORD] and identify which of these questions have the highest search volume and weakest existing answers in the top 10 results.
Then:
1. Select the 5–7 strongest questions to include
2. Write a 2–4 sentence answer for each, in the voice defined in @[STYLE_GUIDE]
3. Cite specifics from the article body where the answer is already covered, so the FAQ links into the rest of the piece
Variables to fill in:
- [TOPIC] — The article’s subject, e.g., Kubernetes monitoring
- [PASTE_QUESTIONS_HERE] — Paste the list you exported from AlsoAsked or Answer the Public
- [PRIMARY_KEYWORD] — Same as Prompt 1
What to expect: Concise answers that read like a human wrote them, with each question chosen because it has search demand and weak coverage in the existing SERP. If the answers feel like restatements of the article, prompt Cursor to lead with the answer first and link to the section for depth.
Follow-up prompt:
The questions are good but the answers are too long and read like blog paragraphs. Rewrite each answer in 2–3 sentences, lead with the direct answer in the first sentence, and cut any throat-clearing.
Prompt 4: Identify related topic clusters to expand coverage
Best for: Finding adjacent topics that should be covered in this article or spun out into new ones. One of the prompts Anish reuses across most refreshes.
Use with: Cursor (with DataForSEO MCP connected)
Prompt 4
For @[ARTICLE_FILE] targeting [PRIMARY_KEYWORD]:
Using the DataForSEO MCP, identify keyword clusters and related topics that:
1. Have meaningful monthly search volume
2. Are not currently covered in the article
3. Are within scope of the article’s topic (do not suggest tangents)
Output a table with: cluster name, representative keyword, monthly search volume, current article coverage (yes / partial / none), and a recommendation (expand existing section / add new section / spin out into a separate article).
Be honest about gaps. If a cluster does not belong in this article, say so.
Variables to fill in:
- [ARTICLE_FILE] — The article you’re refreshing
- [PRIMARY_KEYWORD] — The article’s target keyword
What to expect: A table that maps real search demand against your existing coverage, with clear recommendations. Use this output to plan the rest of the refresh and to seed your next content brief.
Follow-up prompt:
Take the clusters you marked “expand existing section” and draft a 1‑paragraph addition for each, in the voice of @[STYLE_GUIDE]. Show me the suggested insertion point in @[ARTICLE_FILE] for each one.
Prompt 5: Final pass to remove AI tone and filler
Best for: The cleanup step before you route the draft to a subject matter expert.
Use with: Cursor
Prompt 5
Read the full updated @[ARTICLE_FILE].
Strictly follow @[STYLE_GUIDE].
Find and rewrite anything that reads as AI-generated, including:
- Generic openers (“In today’s fast-paced world…”)
- Inflated adjectives (“comprehensive,” “robust,” “seamless,” “cutting-edge”)
- “It’s not X. It’s Y.” constructions
- Negative parallelisms (“No X. No Y. Just Z.”)
- Vague closers (“The possibilities are endless”)
- Filler words like “just” used as a minimizer
Preserve every technical claim, command, code block, and version number exactly as written. Show me a diff of what you changed.
Variables to fill in:
- [ARTICLE_FILE], [STYLE_GUIDE] — Same files referenced throughout
What to expect: A cleaner draft with the AI tells removed. The diff lets you spot any technical content Cursor changed by accident before you accept the edits.
Follow-up prompt:
You changed the technical accuracy in [SPECIFIC_SECTION]. Revert that change and only edit for tone and filler going forward. Do not touch commands, code, version numbers, or numbered specifications.
Tips for better results
- Keep your MCP server list lean: Anish only runs DigitalOcean MCP and DataForSEO MCP. More connected tools means more chances for the AI to wander.
- Tag files like you tag people: Reference briefs and style guides with @filename so Cursor knows which documents to treat as ground truth.
- Refuse to refresh a whole article in one prompt: Section by section produces better drafts and gives you a chance to calibrate as you go.
- Verify the live data is actually flowing: Watch for successful calls to the search volume API or organic live API in the action log. If you don’t see them, the prompt likely fell back on training data.
- Always close with a human pass: As Anish puts it, “AI is a great tool, but it still needs an expert operator. It needs somebody to steer the ship.”
A step-by-step checklist for you
Get Anish’s checklist to build a Cursor workspace wired to live search data through Model Context Protocol (MCP) servers, with editorial context loaded as project files, ready to refresh stale articles section by section
Related
More data from the AI Lab.