How to Use ChatGPT 5.5 to Rewrite Meta Descriptions in Bulk

The Meta Description Problem You Actually Have

Your meta descriptions are invisible until someone searches for your content. Then they appear in Google's search results for maybe two seconds. That's the entire window you have to convince someone to click your link instead of a competitor's.

Most founders don't touch meta descriptions after launch. They're boring. They're technical. They feel like they don't matter. But here's the brutal truth: Google frequently rewrites your meta descriptions with AI-generated text, which means the ones you wrote might not even show up. When Google doesn't rewrite them, your meta description directly impacts click-through rate (CTR). A 2% improvement in CTR is a 2% improvement in organic traffic for the same rankings.

If you have 100 pages indexed and each one gets 10 monthly searches, that's 1,000 potential clicks. A bad meta description might convert 20% of those. A great one converts 30%. That's 100 extra clicks per month—completely free.

The problem: rewriting 100 meta descriptions manually takes four hours. You don't have four hours. You have 20 minutes.

ChatGPT 5.5 changes this equation. You can rewrite your entire meta description library in minutes, not days. But you need the right prompt, the right workflow, and a QA process that catches the AI's mistakes before they hit production.

This guide walks you through the exact system we use at Seoable to bulk-rewrite meta descriptions for technical founders and indie hackers who don't have agency budgets.

Prerequisites: What You Need Before You Start

Before you open ChatGPT, gather these three things:

1. Access to ChatGPT 5.5 (or a newer model)

You need a ChatGPT Plus subscription ($20/month) or ChatGPT Pro ($200/month). Free tier won't cut it—you need the latest model. ChatGPT 5.5 has better instruction-following and fewer hallucinations than older versions. If 5.5 isn't available when you read this, use the latest available model.

2. A spreadsheet of your current meta descriptions

Export your meta descriptions from your site. If you're on WordPress, use a plugin like Rank Math or Yoast to export them. If you're on a custom stack, query your database or use a crawler like Screaming Frog to pull them.

Your spreadsheet should have three columns:

• Column A: URL
• Column B: Page Title
• Column C: Current Meta Description

If you have 50+ pages, add a fourth column:

• Column D: Target Keyword (optional but helpful)

3. Understanding of your target audience and brand voice

ChatGPT can't know what your customers want to see. You need to tell it. Spend 10 minutes writing a one-paragraph brief about your business, who you serve, and what makes you different. Example:

"We're a no-code SEO platform for technical founders. Our customers are indie hackers and bootstrappers who have shipped products but lack organic visibility. They're skeptical of agency jargon and want concrete, measurable results. Our tone is direct, no-nonsense, and credible."

That paragraph becomes part of your prompt. It's the difference between generic AI descriptions and ones that actually convert.

Step 1: Build Your Bulk Rewrite Prompt

This is the single most important step. A bad prompt produces bad descriptions. A good prompt produces descriptions that actually drive clicks.

Here's the prompt template we use. Copy it. Customize it. Use it.

You are an SEO expert writing meta descriptions for [YOUR BUSINESS NAME].

Our business: [PASTE YOUR ONE-PARAGRAPH BRIEF HERE]

Meta description guidelines:
- Length: 150-160 characters (Google cuts off longer ones)
- Include the target keyword naturally (don't force it)
- Start with a benefit or outcome, not a feature
- Use active voice
- Include a reason to click (urgency, specificity, curiosity)
- Avoid generic language like "Learn more" or "Discover"
- Don't duplicate meta descriptions across pages
- Match the searcher's intent

I will give you rows of data in this format:
URL | Page Title | Current Meta Description | Target Keyword (optional)

For each row, rewrite the meta description following the guidelines above. Return ONLY the rewritten meta descriptions in the same order, one per line. Do not include URLs, page titles, or explanations. Just the meta descriptions.

Why this prompt works:

1. It gives context. ChatGPT knows who you are and what matters to your audience.
2. It sets constraints. Character limits, keyword placement, and tone guidelines prevent hallucinations.
3. It specifies output format. "Return ONLY the rewritten meta descriptions" means no extra text, no numbering, no explanations. Just descriptions you can paste directly into your spreadsheet.
4. It prioritizes benefits over features. Most founders write "Learn about our product." ChatGPT will write "Cut SEO costs 80% in 60 days—no agency required."

Don't skip customizing this prompt. Generic prompts produce generic descriptions. Specific prompts produce descriptions that convert.

Step 2: Format Your Data for ChatGPT

ChatGPT has a token limit. If you feed it 500 meta descriptions at once, it will either cut off halfway through or produce garbage at the end. Instead, batch your descriptions into groups of 20-30.

Here's why: A single meta description is roughly 30-40 tokens. Add the URL and page title, and you're at 50-60 tokens per row. Multiply by 25 rows, and you're at 1,500 tokens input. ChatGPT can handle that easily. Go to 50 rows, and you're pushing 3,000 tokens. The model starts to lose focus.

Batch strategy:

• 1-20 pages: One batch
• 21-50 pages: Two batches
• 51-100 pages: Four batches
• 100+ pages: Five batches of 20 each

Format each batch like this:

blog/seo-audit-guide | How to Conduct an SEO Audit | Learn SEO auditing basics. | SEO audit
blog/keyword-research | Keyword Research for Beginners | Find keywords people search for. | keyword research
blog/technical-seo | Technical SEO Checklist | Fix crawl errors and improve indexing. | technical SEO

Use pipe separators (|) to keep columns clean. Paste this into ChatGPT, then paste your prompt above it.

Don't overthink the formatting. ChatGPT is flexible. As long as the data is organized and the prompt is clear, it will parse it correctly.

Step 3: Run the Prompt and Capture Output

Open ChatGPT. Paste your prompt first. Then paste your batch of data.

Wait for the response. It should come back with 20-30 meta descriptions, one per line, in the same order as your input.

Copy the output. Paste it into Column D of your spreadsheet (or Column E if you already have target keywords).

Pro tip: If ChatGPT returns numbered descriptions ("1. Description here"), use Find & Replace to strip the numbers before pasting into your spreadsheet. Highlight the output, open Find & Replace, search for ^\d+\.\s (regex), replace with nothing. Done.

Repeat this process for each batch until all your meta descriptions are rewritten.

Don't close ChatGPT yet. You'll need it for the QA step.

Step 4: The QA Process—Catch AI Mistakes Before Production

Here's where most founders fail. They rewrite 100 meta descriptions and push them live without checking. Then they discover ChatGPT hallucinated a feature that doesn't exist, or wrote a description that's 180 characters instead of 160.

You need a QA process. It takes 30 minutes for 100 descriptions. It saves you from ranking drops.

The QA checklist:

1. Character count (2 minutes)

Add a formula to your spreadsheet to count characters in the new meta description:

=LEN(D2)

Sort by character count. Flag anything over 160. Flag anything under 100 (too short usually means ChatGPT didn't write enough).

If a description is too long, paste it back into ChatGPT with this prompt:

"This meta description is [X] characters. Trim it to 155 characters or less while keeping the core message. Return only the trimmed description."

Copy the trimmed version back into your spreadsheet.

2. Keyword inclusion (3 minutes)

If you have target keywords in your spreadsheet, use Find to check that each keyword appears in its corresponding meta description. Don't force keywords—some descriptions work better without them. But if the keyword is searchable and relevant, it should be there.

3. Sense check (15 minutes)

Read 20-30 descriptions at random. Not all 100. Just a sample. Ask yourself:

• Does this sound like our brand?
• Would I click this in a search result?
• Is this accurate? (Did ChatGPT invent something?)
• Does it match the page content?

If you find issues, note them. Paste those specific descriptions back into ChatGPT with feedback.

Example: "This description mentions 'free trial' but our product doesn't have a free trial. Rewrite it without mentioning free trial. Keep it under 160 characters."

4. Duplicate check (5 minutes)

Use a pivot table or conditional formatting to find duplicate descriptions. ChatGPT sometimes repeats itself, especially if you batch-process 50+ pages at once.

If you find duplicates, rewrite them manually or run them through ChatGPT again with a note: "This description is too similar to [other description]. Make it more specific to [page topic]."

5. Tone and voice check (5 minutes)

Read a few descriptions. Do they match your brand voice? If your brand voice is irreverent and direct (like Seoable), and ChatGPT wrote something corporate and generic, rewrite it.

Paste the generic one back into ChatGPT: "This description is too corporate. Make it more direct and skeptical. Our customers are founders who hate marketing jargon. Rewrite it in that voice."

The brutal truth: You will find mistakes. ChatGPT hallucinates. It invents features. It forgets constraints. That's why QA exists. 20 minutes of QA catches 90% of problems before they hit production.

Step 5: Import Your New Meta Descriptions

Once QA is done, you have a spreadsheet of rewritten, checked, approved meta descriptions.

Now you need to get them into your site.

If you're on WordPress:

If you used Rank Math to export, you can import them back the same way. Export your spreadsheet as CSV. Go to Rank Math → Tools → Bulk Editor. Upload your CSV. Map the columns (URL, Meta Description). Click Import.

If you used Yoast, the process is similar. Go to SEO → Tools → Bulk Editor. Upload CSV. Import.

If you're not using a plugin, you'll need to manually update each page or use a custom script. Ask your developer to write a script that takes a CSV and updates the post_excerpt field in your WordPress database for matching post_names. It's a 20-minute job for any developer.

If you're on a custom stack:

Work with your developer. Give them the CSV. They'll write a script to update your meta description field. It's straightforward—just a database update based on URL matching.

If you're on Shopify, Webflow, or another platform:

Check if they have bulk editing tools. Most do. If not, you're updating manually or hiring someone to build an import script.

Once imported, verify in your staging environment. Click a few random pages. Check that the meta descriptions updated correctly. Then push to production.

Step 6: Monitor Impact and Iterate

You've rewritten your meta descriptions. Now what?

You need to measure whether they actually improved CTR. This is where most founders stop. They don't track results. They don't know if the rewrite worked.

Set up tracking:

1. Baseline measurement (today)

Go to Google Search Console. Click Performance. Note your current CTR. Write it down. This is your baseline.

2. Weekly check-in (every Thursday)

Check GSC again. Is CTR trending up? Down? Flat? It takes 2-3 weeks for changes to show up in the data, so don't panic if nothing changes week one.

3. Monthly review (day 30)

After 30 days, compare your CTR to baseline. If you went from 2.8% to 3.2%, that's a 14% improvement. If you have 10,000 monthly searches, that's 40 extra clicks per month.

If CTR improved, you're done. Celebrate. Move on to other SEO work.

If CTR stayed flat or dropped, something's wrong. Investigate:

• Did Google rewrite your descriptions? (Check GSC—Google shows what snippet it's actually displaying)
• Are the descriptions matching search intent? (Click a few of your results. Does the page match what the description promises?)
• Are your rankings the same? (If rankings dropped, that's the problem, not the descriptions)

If you need to iterate, pick the worst-performing 10-20 descriptions and rewrite them again using this same process.

But here's the thing: most meta descriptions get rewritten by Google anyway. So if you see no change, it might not be because your descriptions are bad. It's because Google decided to use AI-generated summaries instead. In that case, focus on making your page content itself more compelling—Google's AI summaries come from your actual content.

Pro Tips and Common Mistakes

Pro tip #1: Use batching to avoid token limits

Don't paste 200 meta descriptions at once. ChatGPT will cut off or produce garbage at the end. Batch in groups of 20-30. It's faster and more reliable.

Pro tip #2: Include your target keyword in the prompt, not just the data

If you have specific keywords you want to rank for, tell ChatGPT upfront: "Our main keywords are: [keyword 1], [keyword 2], [keyword 3]. Include these naturally in the descriptions where relevant."

Pro tip #3: Rewrite your worst performers first

Don't rewrite all 100 descriptions at once. Start with the 20 pages that get the most search traffic but have the lowest CTR. Those 20 rewrites will have the biggest impact.

You can see this in Google Search Console. Go to Performance. Sort by Impressions (descending). Then look at CTR. The pages with high impressions and low CTR are your priority.

Pro tip #4: Use the same prompt for consistency

Once you've built a prompt that works, use it for all your batches. Don't change it mid-process. Consistency matters. If you change the prompt for batch three, you'll get descriptions that don't match batches one and two in tone and style.

Pro tip #5: A/B test your approach

Rewrite 20 descriptions. Push them live. Wait two weeks. Check CTR. If CTR improved, rewrite the rest. If it didn't, figure out why before wasting time on the other 80.

Common mistake #1: Writing descriptions that are too long

Google cuts off meta descriptions at 160 characters on desktop, 120 on mobile. If you write 200-character descriptions, Google truncates them. ChatGPT knows this if you tell it. Include "Length: 150-160 characters" in your prompt. Then QA to verify.

Common mistake #2: Forcing keywords

Don't write: "Best SEO audit tool for SEO audits to audit your SEO." That's keyword stuffing. It doesn't convert. Tell ChatGPT: "Include the target keyword naturally, but only if it fits. Don't force it."

Common mistake #3: Not matching search intent

If someone searches "how to do X," your meta description should promise to teach them how to do X. If your description says "Buy our tool," they won't click. ChatGPT will get this right if you tell it: "Match the searcher's intent. If the page is a how-to guide, the description should promise step-by-step instructions."

Common mistake #4: Skipping QA

I mentioned this earlier. Don't skip it. 30 minutes of QA saves you from pushing broken descriptions live. ChatGPT hallucinates. QA catches it.

Common mistake #5: Not tracking results

You rewrite descriptions. You should measure whether they worked. Set a reminder to check Google Search Console in 30 days. If you don't track, you don't know if this was worth your time.

Workflow Optimization: Doing This Faster

If you have 500+ pages, this process takes a while. Here's how to speed it up:

Use a ChatGPT API script instead of the web interface

If you're technical, you can write a Python script that:

1. Reads your CSV
2. Batches descriptions into groups of 25
3. Sends each batch to ChatGPT API
4. Captures responses
5. Writes results back to CSV

This runs in parallel. Instead of waiting for 20 batches sequentially (20 minutes), you can run 5 batches at once (4 minutes).

Here's a rough template:

import openai
import csv

openai.api_key = "your-api-key"

with open('meta_descriptions.csv') as f:
    rows = list(csv.reader(f))

batch_size = 25
for i in range(0, len(rows), batch_size):
    batch = rows[i:i+batch_size]
    prompt = "[YOUR PROMPT HERE]\n\n"
    for row in batch:
        prompt += f"{row[0]} | {row[1]} | {row[2]}\n"
    
    response = openai.ChatCompletion.create(
        model="gpt-4-turbo",
        messages=[{"role": "user", "content": prompt}]
    )
    
    # Process response and write to output CSV

If you're not technical, don't bother. The web interface is fine. 20 minutes is fast enough.

Automate QA with a second prompt

After ChatGPT generates descriptions, run them through a QA prompt:

Review these meta descriptions for SEO best practices:
1. Are they 150-160 characters?
2. Do they include the target keyword naturally?
3. Are they duplicate-free?
4. Do they match the brand voice?
5. Are there any factual errors or hallucinations?

Return a list of descriptions that need revision, with the reason why.

[PASTE DESCRIPTIONS HERE]

This catches 80% of issues automatically. You still need to spot-check, but it saves time.

Connecting This to Your Broader SEO Strategy

Meta description rewriting is a single tactical win. It improves CTR. But it's part of a larger SEO system.

If you want to move faster, you need to understand how meta descriptions fit into the bigger picture.

Start with a domain audit to understand what's actually holding back your organic visibility. Is it crawl errors? Thin content? Missing keywords? Bad meta descriptions? Probably multiple things.

Once you know what to fix, you can prioritize. Meta descriptions might be worth 20% of your SEO effort. Content might be 50%. Technical fixes might be 30%.

If you're building an SEO system from scratch as a founder, check out the busy founder's AI stack for SEO. It walks you through the minimal set of tools and processes you actually need—no bloat, no agency overhead.

You should also understand how to set up Google Search Console in 10 minutes if you haven't already. GSC is where you'll track whether your meta description rewrites actually worked.

And if you're serious about SEO, learn how to use ChatGPT for SEO beyond just meta descriptions. ChatGPT can help with keyword research, content briefs, title tags, schema markup, and more.

For a complete step-by-step system, check out the busy founder's brief template for AI-generated content. It shows you how to create briefs that produce ranking content in minutes—not just meta descriptions, but full articles.

If you want to go deeper on technical SEO, adding FAQ schema to your site without touching code is a high-ROI task that ChatGPT can help with too.

And once you've built some content and fixed your meta descriptions, you'll want to set up rank tracking on a bootstrapper's budget to monitor whether your efforts are actually moving the needle.

The Reality Check: Will This Actually Work?

Let's be honest. Meta descriptions don't rank pages. Content does. Links do. Technical SEO does. Meta descriptions improve CTR. That's it.

If you're ranking #5 for your target keyword, a great meta description might bump you from 1% CTR to 2% CTR. That's a 100% improvement in clicks from that keyword. But you're still at #5. You're still getting fewer clicks than the #1 result.

So when should you rewrite meta descriptions?

Rewrite them if:

• You're already ranking for keywords (top 10 results)
• Your current CTR is below 2% (it should be 2-3% for top 10 results)
• You have 50+ pages indexed
• You've already fixed major crawl errors and content gaps

Skip them if:

• You have fewer than 20 pages indexed
• You're not ranking for anything yet
• Your site has major technical issues
• You don't have a content strategy

Meta descriptions are a multiplier, not a foundation. You need the foundation first.

That said, research from Seer Interactive on using ChatGPT to rewrite meta descriptions shows mixed results. Some descriptions improved CTR. Some decreased it. The difference was in the quality of the prompt and the QA process.

This guide gives you both. Follow it, and you'll see improvements. Skip the QA, and you might see decreases.

Summary: The Exact Workflow You Need

Here's the process in one paragraph:

Export your meta descriptions into a spreadsheet (Column A: URL, Column B: Title, Column C: Current Description, Column D: Keyword). Write a custom prompt that includes your brand voice and meta description guidelines. Batch your descriptions into groups of 20-30. Paste your prompt + batch into ChatGPT. Copy the output back into your spreadsheet. Run QA: check character counts, verify keyword inclusion, sense-check for accuracy, flag duplicates, and check tone. Fix any issues by re-prompting ChatGPT. Once all descriptions are approved, import them into your site using your CMS's bulk editor or a custom script. Wait 30 days. Check Google Search Console. Compare CTR to baseline. If CTR improved, you're done. If not, investigate why and iterate.

Total time: 2-3 hours for 100 pages. Total cost: $0 if you already have ChatGPT Plus, or $20/month if you don't.

Total impact: 10-20% improvement in CTR for your indexed pages. If you have 10,000 monthly searches, that's 100-200 extra clicks per month. Completely free.

Key Takeaways

1. Meta descriptions don't rank you, but they improve CTR. A 1% CTR improvement on 10,000 monthly searches is 100 extra clicks per month.

2. ChatGPT 5.5 can rewrite 100 meta descriptions in 20 minutes. But only if you give it the right prompt and batch your data correctly.

3. The prompt is everything. Generic prompts produce generic descriptions. Specific prompts that include your brand voice, audience, and constraints produce descriptions that convert.

4. QA is non-negotiable. ChatGPT hallucinates. 30 minutes of QA catches mistakes before they hit production.

5. Track your results. Check Google Search Console in 30 days. If CTR improved, you've won. If not, investigate why and iterate.

6. Meta descriptions are a tactical win, not a strategy. Do this after you've fixed crawl errors, built content, and established keyword targets. It's a multiplier, not a foundation.

7. Batch processing is faster and more reliable than processing all descriptions at once. Groups of 20-30, not 100+.

You now have the exact workflow to rewrite your meta descriptions without an agency. Use it. Track the results. Move on to the next SEO task.