Generate Instagram captions from hashtag analysis using Apify + GPT-4o-mini

This n8n workflow scrapes recent Instagram posts by hashtag and generates new, relevant caption ideas using OpenAI. It avoids making up suggestions by analyzing real-world content and surfacing common patterns.

---

✅ Use Case

Marketing teams, content creators, or social media managers can:

• Discover what’s trending for a specific topic
• Automatically generate Instagram captions based on real posts
• Understand common caption styles for a niche
• Save time brainstorming ideas while staying on-brand

---

🧠 How It Works

1️⃣ Manual Trigger

🧩 Node: When clicking ‘Execute workflow’

• Manually starts the workflow for testing or single-run execution.

---

2️⃣ Define the Hashtag

🧩 Node: Create Search Term

• Sets the value of the hashtag you'd like to scan.
• Default is n8n, but you can modify it to anything.

{
  "Search_Term": "yourCustomHashtag"
}

---

3️⃣ Scrape Instagram Posts

🧩 Node: Find Recent Posts

• API: Apify Instagram Hashtag Scraper
• Setup:
  • Visit Apify Console
  • Create an API token
  • In n8n, go to Credentials and add HTTP Query Auth
  • Use ?token=yourTokenHere as the query string

JSON Body:

{
  "hashtags": ["{{ $json.Search_Term }}"],
  "resultsLimit": 20,
  "resultsType": "posts"
}

---

4️⃣ Extract Captions

🧩 Node: Set bio and follower count

• Extracts just the caption from each post and stores it in a clean variable for the AI agent to use.

---

5️⃣ Aggregate Captions

🧩 Node: Aggregate

• Gathers all captions into one list before processing.
• Useful for passing a large text block into the AI.

---

6️⃣ Convert to Single Text Block

🧩 Node: Convert table names and columns into single text for agent

• Uses a Code node to combine all captions into a single string for OpenAI to read:

return [
  {
    json: {
      text: items
        .map(item => `- ${JSON.stringify(item.json)}`)
        .join('\n\n'),
    },
  },
];

---

7️⃣ Generate Caption Ideas with AI

🧩 Node: AI Agent

• Takes the combined post text and sends it to GPT-4o-mini.
• Includes this system message:

I'm looking for ideas for posts about {{ $('Create Search Term').item.json.Search_Term }}.
Here’s the last 5 posts on Instagram about the topic. Use those to help me generate a list of relevant captions.

Do not make up ideas that are not like the others in the list.

Output like this:
{
  "Post Idea": ["Idea1", "Idea2"],
  "Most Common Post": ["common post 1", "common post 2"]
}

---

8️⃣ Choose Language Model

🧩 Node: OpenAI Chat Model

• Model: gpt-4o-mini
• Credential: Use your OpenAI API key.
  • Get it from: OpenAI API Keys
  • Add it in n8n under OpenAI credentials.

---

9️⃣ Parse the AI Output

🧩 Node: Structured Output Parser

• Parses the GPT response into structured JSON:

{
  "Post Idea": ["Idea1", "Idea2"],
  "Most Common Post": ["common post 1", "common post 2"]
}

---

🔟 Split the Outputs

🧩 Nodes: Split Out, Split Out1

• Separates the Post Idea list and Most Common Post list into individual items.

---

🔁 Merge for Final Output

🧩 Node: Merge

• Combines the two split lists into one output stream.

---

👤 Need More Help?

Robert Breen
Automation Consultant | AI Workflow Designer | n8n Expert
📧 [email protected]
🔗 LinkedIn