AI Summaries & Auto-Tagging (NewsSync PRO)

🎯 What it does

AI Summaries automatically enhance your imported RSS content with:

  • ✅ Engaging excerpts — AI-written summaries perfect for listings and social sharing
  • ✅ TL;DR boxes — Optional summary boxes at the top of posts (saves readers time)
  • ✅ Smart tags — Auto-generated keywords/tags for better organization and SEO
  • ✅ Multilingual support — Works with content in any language
  • ✅ Cost-effective — Caching reduces API calls (summaries stored for 7 days)

Before vs After AI

Without AIWith AI Summaries
Generic RSS excerpt (often truncated)Professional, engaging excerpt
Manual tagging (time-consuming)Auto-tags with relevant keywords
Inconsistent formattingUniform, branded content
5-10 minutes per postFully automatic

🤖 Choose Your AI Provider

NewsSync PRO supports 3 AI providers — pick the one that fits your needs:

Option 1: OpenAI (Recommended for quality)

Best for: Production sites, high-quality summaries

Pros:

  • ✅ Best quality (GPT-4o, GPT-4o-mini)
  • ✅ Reliable, battle-tested
  • ✅ Fast responses (1-2 seconds)
  • ✅ Great multilingual support

Cons:

  • ❌ Paid service ($0.15-$5 per 1M tokens)
  • ❌ Requires OpenAI account + API key

Cost estimate:

  • ~500 summaries/month = $2-5/month (using GPT-4o-mini)

How to get started:

  1. Sign up at OpenAI Platform
  2. Add payment method (minimum $5)
  3. Generate API key → Settings → API Keys
  4. Copy key to NewsSync PRO → AI Features → OpenAI API Key

Option 2: GitHub Copilot (Best for free usage)

Best for: GitHub users, developers, testing

Pros:

  • ✅ FREE with GitHub Copilot subscription (or $10/month standalone)
  • ✅ Good quality (GPT-4o models via Azure)
  • ✅ No separate billing/tracking needed
  • ✅ Same OpenAI models under the hood

Cons:

  • ❌ Requires GitHub account
  • ❌ Slightly slower than direct OpenAI (Azure routing)

Who has free access:

How to get started:

  1. Verify you have GitHub Copilot access
  2. Generate Personal Access Token → GitHub Settings → Developer Settings
  3. Scopes needed: copilot (or read:user minimum)
  4. Copy token to NewsSync PRO → AI Features → GitHub Token

Option 3: Custom Provider (Advanced users)

Best for: Claude, Groq, OpenRouter, self-hosted models

Pros:

  • ✅ Use any OpenAI-compatible API
  • ✅ Access to Claude (better for some content types)
  • ✅ Cost optimization (cheaper providers)
  • ✅ Full control (self-hosted models)

Cons:

  • ❌ Requires technical setup (API proxy for some)
  • ❌ Quality varies by provider/model

Popular custom options:

  • Anthropic Claude (via Cloudflare Worker) — Best summarization quality
  • Groq — Fastest responses (sub-second)
  • OpenRouter — Access to all models (Claude, Gemini, Llama)
  • Hugging Face — Free tier available (not recommended for production)

📖 Full setup guide: Custom AI Providers

🚀 Quick Setup Guide

Step 1: Choose your provider

In WordPress Admin → NewsSync PRO → Settings → AI Features:

  1. Preferred AI Provider: Select OpenAI, GitHub Copilot, or Custom
  2. Enter your API key/token in the appropriate field
  3. Click “Test API Key” to verify connection

Step 2: Enable features

Toggle on the features you want:

  • ☑️ AI-Powered Summaries — Generate excerpts and TL;DR
  • ☑️ Show AI Summary Box — Display TL;DR at top of posts
  • ☑️ AI Auto-Tagging — Create/assign tags automatically

Step 3: Configure model (optional)

For OpenAI:

  • Recommended: gpt-4o-mini (fast, cheap, good quality)
  • Best quality: gpt-4o (2x cost, 10% better)
  • Budget: gpt-3.5-turbo (cheapest, lower quality)

For GitHub Copilot:

  • Use default model (handles automatically)

For Custom:

  • Enter your provider’s model ID (e.g., claude-3-5-sonnet-20241022)

Step 4: Save and test

  1. Click Save Changes
  2. Import a test RSS feed item
  3. Check the created post:
    • Excerpt: Should have AI-generated summary
    • Content: TL;DR box at top (if enabled)
    • Tags: Auto-generated keywords (if enabled)

✅ Success! AI is now enhancing all imports automatically.

⚙️ Advanced Settings

AI Calls Per Run

Controls how many items get AI summaries per import batch.

Default: 5 items per batch

Recommendations:

  • Low-volume feeds (1-5 posts/day): Set to -1 (unlimited)
  • High-volume feeds (50+ posts/day): Set to 10-20 (control costs)
  • Testing: Set to 2-3 (verify quality before scaling)

Cache Duration

AI summaries are cached to avoid duplicate API calls.

Default: 7 days

When to adjust:

  • Shorter (1-3 days): If your feed content updates frequently
  • Longer (30 days): For static/archive content (save costs)

Custom Prompts (Advanced)

Edit the AI prompt to change output style/format.

Default prompt:

Summarize the following article in 2-3 sentences.
Then provide 5 relevant keywords.

Title: {title}
Content: {content}

Example customizations:

  • Add “Write in Portuguese” for localized summaries
  • Add “Focus on technical details” for developer content
  • Add “Use simple language for beginners” for tutorials

💰 Cost Optimization Tips

1. Use GPT-4o-mini (not GPT-4)

Savings: 15x cheaper, 95% same quality

2. Limit AI Calls Per Run

Set to 10-20 for high-volume feeds (don’t process every item)

3. Enable caching (default ON)

Summaries stored 7 days → no repeat API calls

4. Use shorter excerpts

Shorter input = lower token cost (adjust in Display settings)

5. Consider GitHub Copilot

If you already have subscription → effectively FREE

Cost comparison (per 1000 summaries):

ProviderModelCost
OpenAIGPT-4o-mini~$0.30
OpenAIGPT-4o~$5.00
GitHub CopilotIncluded$0 (with subscription)
Claude (custom)Sonnet 3.5~$3.00
Groq (custom)Mixtral~$0.10

🔧 Troubleshooting

“Test API Key” fails

OpenAI errors:

  • ❌ Invalid API key → Check key copied correctly (no spaces)
  • ❌ Rate limit exceeded → Wait 60 seconds, try again
  • ❌ Insufficient quota → Add payment method in OpenAI dashboard

GitHub errors:

  • ❌ Unauthorized → Token needs copilot scope (regenerate)
  • ❌ No Copilot access → Verify subscription active
  • ❌ Token expired → GitHub tokens expire (set no expiration)

Custom provider errors:

  • ❌ Timeout → Increase timeout or use faster provider
  • ❌ Bad format → Verify API is OpenAI-compatible

AI summaries not showing in posts

Check:

  1. ✅ AI Features toggle is ON
  2. ✅ “AI Calls Per Run” is not zero
  3. ✅ Post was imported AFTER enabling AI (re-import to test)
  4. ✅ Check WordPress Debug Log for errors

Force re-generate:

  • Delete post → Re-import feed item
  • Or clear transient cache: wp transient delete newssync_ai_*

Summaries are low quality / gibberish

Causes:

  • Using weak model (GPT-3.5, Mistral 7B)
  • Feed content is non-standard (images only, minimal text)
  • Prompt needs customization for your content type

Fixes:

  • ✅ Upgrade to GPT-4o-mini or Claude
  • ✅ Test with different feed (verify it’s not feed-specific)
  • ✅ Customize prompt to match your content

Tagging creates too many/wrong tags

Causes:

  • AI being too creative with keywords
  • Model not trained on your niche

Fixes:

  • ✅ Edit prompt: Add “Use only existing WordPress tags” or “Limit to 3-5 tags”
  • ✅ Use tag blacklist/whitelist (in PRO settings if available)
  • ✅ Manually review first 10 imports, adjust prompt

📊 What Gets Sent to AI

Privacy note: NewsSync sends these fields to your AI provider:

  • Article title
  • Article excerpt/content (truncated to ~500 words)
  • Your custom prompt

NOT sent:

  • User data, emails, passwords
  • WordPress admin data
  • Other posts/content

Cache: Results cached locally (WordPress database) for 7 days.

🆘 Need Help?

Support Channels

Pro Tip: Start with GitHub Copilot (free!) or OpenAI’s GPT-4o-mini (cheap). Test with 10-20 posts before enabling site-wide. Monitor costs in your provider dashboard. 🚀

Back to Documentation