← Back to Index

← Back to Index

Solutions • Announce Funding • Build Brand Awareness • Generate More Leads • Reduce Cost per Lead • Launch a New Product • Be Found Online • Services • Public Relations • Growth Marketing • PPC • GEO • SEO • ABM • Content Marketing • Email Marketing • Sales Enablement • Thought Leadership Content • Brand Awareness Content • Multiplier Marketing • Audits • Content Marketing • Paid Media • SEO • Martech Configuration • Clients • Clients • Case Studies • Client Testimonials • Resources • Resource Center • Blog • Podcast • Deep Dives • Store • Newsletter • About • About Us • Why Firebrand • Team • Join • Values • Our Pledge • Contact

Putting ChatGPT for SEO Metadata to the Test - Firebrand

Since ChatGPT’s explosion of growth starting in 2022, it has become the go-to free tool for generative AI across many industries — with digital marketing being one the most widely adopted. Within digital marketing, the use of ChatGPT for SEO has grown tremendously and ChatGPT SEO use cases are becoming more and more refined. Our Firebrand Labs team’s mission is to test generative AI tools and determine where they may fit best for our clients. So, we have put  ChatGPT for SEO to the test — specifically with its ability to help generate metadata efficiently and accurately. Let’s dive into the best ChatGPT prompts for SEO metadata, and how to use the tool to your advantage when executing metadata audits and fixes for your indexable web pages.

How our test worked

Our approach began with identifying key criteria that would be used as a determining factor for ChatGPT’s efficacy as a tool for generating metadata. The end goal was clear: using ChatGPT for SEO metadata generation would have to not only be time-saving, but produce outputs of adequate quality that accurately describe webpage material. In this test, we used ChatGPT for meta descriptions and also did some testing with rewriting and generating title tags.

With that said, the following are key qualifiers we used to identify the best ChatGPT prompts for SEO metadata and its overall effectiveness:

  • \\- ChatGPT’s ability to generate metadata when given a prompt containing multiple web pages
    • \\- ChatGPT’s ability to generate metadata within given character count parameters - ChatGPT’s ability to generate multiple metadata options to choose from for a given web page or set of web pages in a prompt - ChatGPT metadata quality in providing accurate meta descriptions for web pages - ChatGPT’s ability to improve metadata outputs based on \\\_follow up\\\_ prompts

    Read on for a full recap of ChatGPT for SEO metadata:

    The first part of our test focused on testing ChatGPT for meta descriptions. We started out using ChatGPT with the intention of making metadata generation quicker and more efficient for identified web pages than manual writing. Our first ChatGPT SEO prompt tested our first key qualifier: can ChatGPT generate meta descriptions for multiple webpages at once, while also taking into account our character count parameters (that are based on Yoast best practices)?

    Prompt #1: write me a meta description maximum 155 characters for the following webpages:

    https://www.firebrand.marketing/2023/08/the-5-biggest-content-marketing-mistakes-startups-make/

    https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/

    https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/

    https://www.firebrand.marketing/2023/07/why-organic-pr-is-an-essential-part-of-your-content-marketing-strategy/

    https://www.firebrand.marketing/2023/06/unlocking-the-soul-of-your-brand/

    ChatGPT’s Response:

    1\\. “Avoid startup content pitfalls. Learn 5 key mistakes for effective strategies. Expert insights for success.”
    2\\. “Exploring PR thread impact. Stay updated on evolving strategies. Insights into modern PR landscape.”
    3\\. “Beware the auto MQL curse. Discover lead engagement tactics. Navigate effective lead qualification.”
    4\\. “Organic PR’s vital role. Boost brand visibility holistically. Insights for a powerful strategy.”
    5\\. “Unlock your brand’s essence. Strategies for authentic development. Elevate your brand’s impact.”

    The initial output checks off both criteria — meta descriptions were generated for all the URLs included in the prompt, and each are under 155 characters. There was one error made by ChatGPT in regards to content accuracy for URL 2, which makes that meta description null and void. Since ChatGPT showed it could handle five URLs in the prompt, we followed up with a prompt that included 20 URLs as a test for metadata audits for large sites, or sites that have a large number of webpages that have missing or trailing meta descriptions.

    Prompt #2: write me a meta description maximum 155 characters for the following webpages:

    https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/

    https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/

    https://www.firebrand.marketing/2023/06/unlocking-the-soul-of-your-brand/

    https://www.firebrand.marketing/about-pr-digital-marketing-agency-san-francisco/

    https://www.firebrand.marketing/case-studies/

    https://www.firebrand.marketing/multiplier-marketing/

    https://www.firebrand.marketing/resources/

    https://www.firebrand.marketing/store/

    https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/

    https://www.firebrand.marketing/contact-bay-area-pr-marketing-firm/

    https://www.firebrand.marketing/2020/12/understanding-the-value-of-assisted-conversions/

    https://www.firebrand.marketing/client-testimonials/

    https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/

    https://www.firebrand.marketing/why-firebrand/

    https://www.firebrand.marketing/safe-start-pledge/

    https://www.firebrand.marketing/firebrand-team/

    https://www.firebrand.marketing/join-best-digital-marketing-agencies-to-work-for/

    https://www.firebrand.marketing/cookie\\\\\\\_policy/

    ChatGPT’s Response:

    1\\. Avoid startup content pitfalls. Learn 5 key mistakes for effective strategies. Expert insights for success.”
    2\\. “Exploring PR thread impact. Stay updated on evolving strategies. Insights into modern PR landscape.”
    3\\. “Beware auto MQL curse. Discover lead engagement tactics. Navigate effective lead qualification.”
    4\\. “Organic PR’s vital role. Boost brand visibility holistically. Insights for a powerful strategy.”
    5\\. “Unlock brand’s essence. Strategies for authentic development. Elevate your brand’s impact.”
    6\\. “About us – PR & digital marketing agency in SF. Crafting success stories with strategic expertise.”
    7\\. “Success stories that speak. Explore our case studies for insights into marketing triumphs.”
    8\\. “Unlock growth with Multiplier Marketing. Amplify results through strategic collaboration.”
    9\\. “Fuel your insights with our resources. Stay informed and empowered for marketing success.”
    10\\. “Discover our store. Resources to drive your marketing journey. Elevate your strategies.”
    11\\. “Get insights on PR threads’ significance. Navigate evolving PR landscape with our expertise.”
    12\\. “Connect with us. Leading Bay Area PR & marketing firm. Let’s create success together.”
    13\\. “Understanding assisted conversions’ value. Unveil the impact on your marketing strategies.”
    14\\. “Hear from our clients. Real stories of success. Partner with us for your marketing journey.”
    15\\. “Explore auto MQL curse’s impact. Strategize lead engagement for better conversions.”
    16\\. “Why choose Firebrand? Discover our approach to PR & marketing success.”
    17\\. “Our commitment to safety. Explore our Safe Start Pledge for a secure partnership.”
    18\\. “Meet the Firebrand team. Experts driving PR & marketing excellence.”
    19\\. “Join the best digital marketing agencies. Be a part of our innovative team.”
    20\\. “Our cookie policy. Learn how we use data to enhance your browsing experience.”

    The expanded URL test was successful here as well. ChatGPT generated meta descriptions for all 20 URLs included in the prompt, and each are under 155 characters.

    Now that we know a large batch of URLs can be entered within a prompt to generate ChatGPT metadata, it’s time to assess quality further as a key indicator. In an attempt to objectively assess meta description quality, we used the following qualification questions:

  • \\- Did the ChatGPT meta description accurately reflect the purpose of the page?
    • \\- Did the ChatGPT meta description provide enough information about the page? - Did the ChatGPT meta description provide information that matches the form of webpage content? (i.e does it accurately convey the webpage as a thought leadership piece, core page etc.) - Is manual editing needed (and if so, how much) to finalize a ChatGPT meta description before it goes live on a website?

    With these key qualifiers in place, we examined the ChatGPT meta description outputs — where we found mixed results.

    Let’s first look at one of the better examples: for the page “https://www.firebrand.marketing/client-testimonials/” the meta description in output #14 reads “Hear from our clients. Real stories of success. Partner with us for your marketing journey.” Here the ChatGPT meta description checks off all boxes, where the purpose of the page is conveyed, accurately described, and is good enough to require no manual edits before it can go live.

    Now, let’s look at a subpar example: for the page “https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/ the meta description in output #15 reads “Explore auto MQL curse’s impact. Strategize lead engagement for better conversions.” Here the ChatGPT meta description fails in a few areas. First off, the ChatGPT-generated meta description does not do an adequate job conveying that this is a thought leadership piece with vague messaging and wording. The meta description fails to either highlight a key takeaway from the blog piece or convey that it is an informative piece of content. The use of multiple verbs to start phrases confuses the reader that it could be webpage content with a service offering. Additionally, manual editing is necessary to finalize this meta description before it goes live on the website.

    These mixed results created the need for an additional test

    We examined ChatGPT’s ability to generate multiple metadata options for a set of web pages in a given prompt. This should provide us with enough options that one is adequate so we can reduce manual editing as much as possible. Because let’s not forget, the end goal of this generative AI test is to save time in making metadata changes.

    Prompt #3: write me 3 meta descriptions maximum 155 characters for each of the following webpages:

    https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/

    https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/

    https://www.firebrand.marketing/2023/06/unlocking-the-soul-of-your-brand/

    https://www.firebrand.marketing/case-studies/

    https://www.firebrand.marketing/multiplier-marketing/

    https://www.firebrand.marketing/resources/

    https://www.firebrand.marketing/store/

    https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/

    https://www.firebrand.marketing/contact-bay-area-pr-marketing-firm/

    https://www.firebrand.marketing/client-testimonials/

    https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/

    https://www.firebrand.marketing/why-firebrand/

    https://www.firebrand.marketing/safe-start-pledge/

    https://www.firebrand.marketing/firebrand-team/

    https://www.firebrand.marketing/cookie\\\\\\\_policy/

    ChatGPT’s Response:

    Startup Content Mistakes:

    1\\. “Steer clear of content pitfalls. Discover startup mistakes for successful marketing strategies.”
    2\\. “Avoiding content errors in startups. Learn from key blunders for effective growth.”
    3\\. “Startup content marketing missteps. Transform mistakes into opportunities for success.”

    Threads & PR Impact:

    1\\. “Unraveling PR thread significance. Stay updated with evolving strategies for modern PR.”
    2\\. “PR threads’ influence explored. Insights into navigating the dynamic PR landscape.”
    3\\. “Discover PR thread impact. Elevate strategies with insights into effective communication.”

    Automatic MQL Challenge:

    1\\. “Beware the auto MQL trap. Engage leads effectively with tactics to overcome challenges.”
    2\\. “Navigating the automatic MQL curse. Strategies to convert leads with meaningful engagement.”
    3\\. “Unlocking auto MQL solutions. Transform challenges into successful lead engagement.”

    \\\_(Note: Due to character limitations, some descriptions may have slight variations in wording while maintaining the essence of the message.)\\\_

    Here, the ChatGPT meta description prompt resulted in three metadata variations as requested, however the response stopped generating after the first 3 URLs in the list. The response did maintain the character count parameters (with a note about the results copy in regards to character limitations). However, the overall goal of generating bulk variations for the entire list of pages was not met. In turn, we followed up and tested a similar ChatGPT meta description prompt with fewer URLs (ten to be exact, half of the previous prompt). The response this time generated a meta description for each webpage but failed to give 3 variations and instead one for each. We followed up with a conversational prompt calling out the error:

    Follow up to Prompt #3: you did not give me 3 variations for each webpage

    ChatGPT’s Response:

    I apologize for the oversight. Here are three variations of meta descriptions for each webpage:

    1\\. https://www.firebrand.marketing/2023/08/the-5-biggest-content-marketing-mistakes-startups-make/:

  • \\- “Don’t stumble on startup content errors. Explore 5 major content marketing mistakes to ensure your success.”
    • \\- “Startup content pitfalls revealed! Learn about the top 5 content marketing mistakes hindering growth.” - “Content marketing missteps startups must avoid. Discover the 5 crucial errors holding back your progress.”

    2\\. https://www.firebrand.marketing/2023/07/will-threads-matter-for-pr/:

  • \\- “Threads and PR: What’s the link? Delve into the significance of threaded discussions for modern PR strategies.”
    • \\- “Unraveling the thread-PR connection. Discover if threaded conversations truly impact effective PR.” - “Threaded discussions in PR: A game-changer? Explore the relevance of threaded content in shaping PR outcomes.”

    3\\. https://www.firebrand.marketing/2023/07/the-curse-of-the-automatic-mql/:

  • \\- “Beware the Automatic MQL curse. Delve into the challenges arising from excessive reliance on automated MQL.”
    • \\- “The downsides of Automated MQL: Exploring its limitations and potential drawbacks.” - “Automated MQL: Blessing or curse? Uncover the complexities tied to automated MQL processes.”

    4\\. https://www.firebrand.marketing/2023/07/why-organic-pr-is-an-essential-part-of-your-content-marketing-strategy/:

  • \\- “Essential role of Organic PR in content strategy. Learn why it’s a must for a holistic approach.”
    • \\- “Organic PR’s vital role in content strategy. Explore why it’s a cornerstone for comprehensive success.” - “Why Organic PR is non-negotiable. Understand its pivotal place in a well-rounded content marketing approach.”

    5\\. https://www.firebrand.marketing/2023/06/unlocking-the-soul-of-your-brand/:

  • \\- “Unleash your brand’s essence. Discover strategies that unveil the true soul of your business.”
    • \\- “Exploring brand essence revelation. Unlock the core identity of your brand with innovative approaches.” - “Revealing your brand’s soul: Strategies to unearth the authentic heart of your business.”

    6\\. https://www.firebrand.marketing/about-pr-digital-marketing-agency-san-francisco/:

  • \\- “Meet Firebrand: Your San Francisco PR & Digital Marketing ally. Learn how we amplify businesses.”
    • \\- “About Firebrand: Elevating brands in San Francisco through PR & Digital Marketing expertise.” - “Empowering brands in San Francisco: Get to know Firebrand, your PR & Digital Marketing companion.”

    7\\. https://www.firebrand.marketing/case-studies/:

  • \\- “Success showcased: Explore Firebrand’s transformative case studies, delivering exceptional results.”
    • \\- “From challenge to triumph: Dive into Firebrand’s case studies highlighting remarkable client journeys.” - “Real-world impact unveiled. Discover how Firebrand’s strategies have driven impressive outcomes.”

    8\\. https://www.firebrand.marketing/multiplier-marketing/:

  • \\- “Experience Multiplier Marketing’s prowess. Elevate your brand with our innovative growth strategies.”
    • \\- “Unlock exponential growth with Multiplier Marketing. Discover the power of amplifying your brand.” - “Amplify your brand’s success: Unleash growth through Multiplier Marketing’s cutting-edge techniques.”

    9\\. https://www.firebrand.marketing/resources/:

  • \\- “Empowering insights await. Access a wealth of resources to stay informed about PR and marketing trends.”
    • \\- “Stay informed with Firebrand’s valuable resources. Your hub for the latest in PR and marketing knowledge.” - “Knowledge is power: Explore our resource center for expert insights on PR and marketing.”

    10\\. https://www.firebrand.marketing/store/:

  • \\- “Elevate your brand with premium solutions. Visit our store for specialized marketing products and services.”
    • \\- “Your brand’s growth journey starts here. Discover our store’s offerings for impactful marketing solutions.” - “Unlock your brand’s potential with our store. Explore a range of products and services for marketing success.”

    Bingo! The follow up to call out the error in the initial ChatGPT meta description prompt resulted in a fix where we were able to obtain three metadata variations for each webpage. We then followed up by testing the initial prompt with the full list of 20 URLs that failed with a conversational call out of the error, but ChatGPT made the same error of only generating meta descriptions for the first few URLs.

    Our takeaway ultimately after testing different quantities of URLs was that ChatGPT could handle generating multiple variations for between 10-15 URLs in a given prompt, but then was not able to correct itself. This could be due to the token limit being reached within GPT 3.5. With that said, we were able to drill down a best practice for the quantity of URLs in a given ChatGPT meta description prompt to successfully output multiple variations, given that our previous testing verified that adequate quality metadata was being generated with the correct parameters in place.

    Takeaways and best practices: ChatGPT for SEO meta descriptions

    Takeaways:

  • ChatGPT copy was better than we had expected overall. With our success metric being efficiency for SEOs, the time-saving aspect could only be achieved if meta descriptions were of adequate quality to reduce manual editing where possible. 
  • The ChatGPT meta description prompt matters. For example, even the placement of where character count parameters entered in a prompt’s text are important, as some tested prompts (not included in this article) generated metadata that was over the given character limits. 
  • ChatGPT is able to generate multiple variations for metadata, when given the right instructions. The generative AI tool can be inconsistent with these results, and the prompts and follow up responses can dictate the outcome.

    • ChatGPT for SEO Metadata Best Practices:

    There are certainly dos and don’ts when it comes to using ChatGPT for SEO metadata. Here is a full breakdown of suggestions from our Firebrand Labs team to get the best results and make metadata audits more efficient:

  • Immediately use prompts to generate multiple metadata variations. As mentioned, the quality of metadata results can vary — so immediately ask for multiple variations per webpage for flexibility and a higher chance of accuracy and quality to avoid manual editing.
  • Be careful how many URLs are entered in a given prompt. Despite fixes to errors with follow up prompts, it seems ChatGPT struggles to generate multiple metadata variations after a certain number of URLs in a given prompt — possibly due to the token limit (4,096) being reached . We found our sweet spot to be ten URLs.
  • Use follow up prompts when errors occur. For example, when ChatGPT only generates one variation for URLs in a given prompt, point out the error in the chat log as we did to obtain the desired output.
  • Don’t reinvent the wheel. Some of our testing has generated the best ChatGPT prompts for SEO, so leverage our research.

    • Overall, after much testing done by Firebrand Labs, we certainly recommend using ChatGPT for SEO metadata. Website metadata audits can take hours (or even days!) depending on the size of a site and how many pages need fixes — whether it’s pages missing metadata or metadata that needs to be corrected or improved. So, leveraging generative AI for metadata copy is certainly becoming an effective way to accelerate this process for SEOs, and when done right it is a no-brainer.

    © 2025 Firebrand Communications LLC

    Original source: https://www.firebrand.marketing/2023/08/putting-chatgpt-for-seo-metadata-to-the-test

    This is an LLM-optimized cache with preserved navigation context and semantic structure.