AI produces better SEO content

🤖 This is a Science Says special of the latest scientific research in AI 🎓 Use it to get better, evidence-based results when using AI in your products and marketing 📈

This insight is brought to you by… Conjointly

Have a gut feeling about a decision?

Test it first, and avoid burning yourself if the hunch is wrong.

Conjointly is the state of the art research platform to help you understand things like:

• Optimize your pricing

• Refine claims and messaging

• Perfect your product offering

Based your decisions on data, not hunches.

Want to sponsor Science Says? Here’s all you need to know.

📝 Intro

One of the most promising and discussed use cases of generative AI (e.g. ChatGPT) for marketers is to produce SEO content (content optimized to rank higher in search engines).

In theory, it should be able to produce good enough quality content, in less time - saving costs.

But does it? Or does the human touch and tone still matter more?

Here’s what the science says.

P.S.: Try prompting your generative AI to use a slightly surprising sentence structure - it makes your message more effective, and increases CTRs and the usefulness of reviews.

Want hundreds more insights like these? Explore all Science Says insights here.

Using AI with a human editor to generate SEO content improves results and cuts costs

Topics: AI | Website & App | Messaging & Copy
For: Both B2C and B2B
Research date: February 2022
Universities: Vienna University of Economics and Business, Emory University and Modul University

📈 Recommendation

Use generative AI (e.g. ChatGPT, Perplexity, or Gemini) to create SEO-optimized website content that’s then reviewed and edited by a human.

Have the AI draft the content based on the current top-ranking pages on a topic, incorporating your brand voice and tone. Then, edit it to improve tone, readability, and correct any mistakes.

Your content will perform better in SEO rankings - with smaller production costs and shorter timelines.

🎓 Findings

• Website SEO content generated through a hybrid approach - with AI drafting the content and humans editing it - achieves better search engine rankings and saves costs and time - compared to content generated entirely by humans.

• As part of 2 field experiments, researchers found that:

  • When comparing 66 content pieces published over 412 days using industry-specific keywords in the IT industry at a midsize international commercial firm and the education industry at a large, internationally recognized public business school:

    • Content generated by AI and revised by an employee (asked to keep changes to a minimum) had an average of 17 pages in the top rankings per day

    • Content made by SEO experts and students trained in SEO had 2 pages each in the top rankings per day

    • Content made by students untrained in SEO had no pages in the top rankings per day

  • When comparing 30 content pieces published over 96 days using industry-specific keywords in the education industry at a large, internationally recognized public business school:

    • AI-generated, human-edited content targeted at industry-specific keywords, like “Master Program in Marketing” outperformed content written by SEO experts both in terms of search engine rankings and the number of pages appearing on the first page of search results (top 10). 

  • On average, the semi-automated content approach resulted in a 91% reduction in labor costs compared to SEO content production entirely by humans.

• There was no significant difference in the readability, credibility, and general attitudes people had towards the content. Both content types were perceived the same way.

🧠 Why it works

• When an AI model generates SEO content, it looks at content that is already ranking highly to generate similar content that performs well.

• Based on the instructions we feed into the AI model, it creates unique content, making sure the content stands out and that the page’s SEO rankings aren’t penalized for having duplicate content.

• We do still need to review and polish the content however to make sure it has a human tone and to edit out any AI hallucinations.

📘 Final countdown: The Playbook of AI is almost here! ⏳

P.S.: This time, on top of following our usual style of Playbooks - which 1,700+ of you absolutely love - we are also asking the top marketing & AI scientists around the world for their practical recommendations of how we should be prepared for what is next in AI.

We’ve already collected some fantastic insights from scientists at business schools like Chicago Booth, Harvard, University of St Gallen, and Bocconi - and many more are coming in. Can’t wait to share them in the playbook!

✋ Limitations

• The research focused on engagement, looking at page rankings, page views, and sessions - it did not look at other user behaviors such as conversion rates or purchases.

• The study looked primarily at the IT services and education sectors. While other product categories weren’t looked at, it’s likely the effect would hold across categories and for highly niche content too.

• The research used ChatGPT-2 to generate the AI content. More advanced models have been released since - and as AI continues to improve, it’s likely the effect will continue to strengthen.

• Training an AI model on the currently top-ranked search results could lead to the creation of content very similar to that already being created, potentially stunting the growth of diverse and original online content.

• As SEO algorithms continue to evolve, they may factor in whether the content is AI or human-generated (if they’re able to detect that), or competition will exponentially increase - reducing the positive impact of using this technique in the future.

• The research focused on Search Engine Optimization - it’s unknown if the same effect would hold for Generative Engine Optimization (optimizing content to be picked up by generative AI models).

🏢 Companies using this

• SEO is a major part of digital marketing efforts for almost all companies, with specialized agencies and services working specifically on optimizing SEO rankings.

• Regular content posting incorporating important keywords, especially through blogs or news and updates sections, are a key way companies drive SEO traffic towards their sites.

• Companies rarely, if ever, admit to using AI to draft their marketing content, possibly due to fears customers will consider this content as generic or not useful.

• AI-powered platforms like Jasper, INK, and Alli AI provide SEO solutions to generate content and optimize the technical SEO of websites.

SEO service Ahrefs lets you check the keywords generating traffic for both your and competitor websites - a key tool to understanding what your content should focus on.

⚡ Steps to implement

• Using a service like Ahrefs or Semrush, track the words people use to search for you and your competitors’ sites online. Based on the search results for these phrases, shortlist the website content that ranks high on these keywords.

• Feed this content into a generative AI model (e.g. Perplexity, ChatGPT) to train it on how to write similar content.

  • In addition to this, you can also upload content related to your brand positioning, the tone and language of your marketing communications, and other key talking points. Remember - the more information you share with your model, the better it will be able to develop content that meets your needs.

• Have your generative AI create new content for your website, including blog posts, updates, and information on product pages. Focus on ensuring that the overall topic, keywords, and readability are similar to the top-ranking content fed into your AI model. You can score the content using Semrush to see how well it uses keywords.

• Have a member of your marketing team review and edit the best-scoring content to ensure accuracy, brand consistency, and a human tone of voice. The more creative and unique you want the content to be, the more extensive the edits should be.

• Once the content has been published, monitor the page’s performance on various SEO keywords to see how well it ranks.

🔍 Study type

Online experiment (588 people in the US) and field experiments (with a midsize IT company tracking website traffic and ranking for 412 days and an international business school tracking website and ranking for 96 days)

📖 Research

Frontiers: Supporting Content Marketing with Natural Language Generation. Marketing Science (February 2022)

🏫 Researchers

• Martin Reisenbichler. Vienna University of Economics and Business

• Thomas Reutterer. Vienna University of Economics and Business

• David A. Schweidel. Emory University

• Daniel Dan. Modul University

Remember: This is a scientific discovery. In the future it will probably be better understood and could even be proven wrong (that’s how science works). It may also not be generalizable to your situation. If it’s a risky change, always test it on a small scale before rolling it out widely.

Here is how else Science Says can help your marketing:

• 📈 Join the Science Says Platform to unlock all 250+ insights, real-world case studies, and exclusive playbooks

• 📘 Boost your sales and profits with topic-specific Science-based Playbooks (e.g. Pricing, Ecommerce, SaaS, AI)

• 🔬 Get on-demand evidence to make better decisions. My team of PhDs and I regularly help leading brands in FMCG (e.g. Mars), retail, and tech. Reach out here.

🎓 It took 3 of us 11.5 hours to accurately turn this 12 page research paper into this 3min insight. 

If you enjoyed it, please share it with a friend. If you’re that friend, you can subscribe below for $0 👇