Marketing to ChatGPT: The Next Evolution of SEO?
A Marketer Influencing the Training of ChatGPT7, Circa 2024

Marketing to ChatGPT: The Next Evolution of SEO?

Have you tried asking ChatGPT about yourself or your company - by name?

If you are Arnold Schwarzenegger, or if you work for Microsoft, chances are you’ll get a result that is spot-on. 

If you’re only moderately famous or work for a small company, chances are you’ll get a result that is “sort of” right, with a few errors (maybe they’ll get your company’s founding year wrong, or where you earned your college degree).

….Or ChatGPT might know nothing about you. She might simply make stuff up (she’s really good at that). 

If you run a small restaurant in Champaign, Illinois  - how do you get ChatGPT to offer your establishment as a recommendation when someone asks “Where’s the best place to eat lasagna near UIUC?”

Or if you’re Kellogg - how do you get ChatGPT to respond to a question about breakfast cereal with “Frosted Flakes? They’re grrrrreat!” 

AIO: The Next Incarnation of SEO

For the past two decades, the evolving discipline of SEO has fought this battle with search engines.

In 2023, a fundamental shift will take place. A new discipline will emerge… Artificial Intelligence Optimization: How to optimize content to increase prominence in the training datasets for AI agents like ChatGPT? 

More specifically:

  • How do you place your content so it’s more likely to be included in the training data?
  • How do you structure your content so it will have the biggest impact on the training algorithm? 
  • How do you write your content to maximize positive associations between your brand/name and the concepts (not keywords) in the artificial neural network?
  • How do you create your content to maximize the likelihood of attention mechanisms focusing on your keywords in the input sequence?

Measurement and Testing

SEO success is measured as follows: in the list of search results for keyword searches, how prominent does your content appear?

AIO success is measured differently. Of course, you can ask subsequently-trained models certain questions and make sure that the answer reflects what you want it to - but there are more objective approaches. You need simply look at the word embeddings for your brand/name to see what other concepts it is associated with. These word embeddings are simply vectors - a list of numbers, each one representing the strength of the association with your brand/name and other concepts. The higher the association value with positive concepts, and the lower the association value with negative concepts, the better you are doing.

Because many models are open-sourced, marketing/AIO agencies will be able to do massively parallel test-runs on various models to see how changes to content affect how the content impacts the neural net. This will actually be fairly easy to implement and not especially computationally intensive, even though these artificial neural networks are usually a ‘black box’. All you need to look at is changes to the weights (referred to as the model’s parameters), and see what the impact is on word embeddings to determine if you’re getting the results you want. 

Best Practices

The rest of this article is made up of the most useful responses by ChatGPT, in response to a barrage of questions on the subject

Where to Place Your Content To Maximize Inclusion in Training Datasets

ChatGPT's training data is sourced from a diverse range of publicly available text found on the internet. There isn't any official guidance on specific locations to place your content for inclusion in ChatGPT's training data. However, by considering some general practices, you can increase the likelihood of your content being included:

  1. High-quality content platforms: Publish your content on popular and authoritative websites, blogs, or platforms related to your niche or industry. Content from reputable sources is more likely to be considered during the data collection process.
  2. Collaborate with influencers or experts: Partner with influencers or experts in your field to increase your content's visibility and credibility. Content from recognizable figures or industry experts may be more likely to be included in training data.
  3. Engage with online communities: Share your content on relevant online communities, forums, or discussion boards, such as Reddit or Quora. These platforms often contain diverse and engaging content, which can make them valuable sources for AI training data.
  4. Social media sharing: Post your content on social media platforms like Twitter, LinkedIn, or Facebook. High-quality content that generates significant engagement may have a higher likelihood of being included in the training dataset.
  5. Open-access repositories: Publish your content in open-access repositories, journals, or platforms that are available to the public and researchers. This type of content is more likely to be accessible and used as a data source for AI models.
  6. Search engine optimization (SEO): Optimize your content for search engines to increase its visibility and discoverability by AI model developers or web crawlers.

While these guidelines don't guarantee your content's inclusion in ChatGPT's training data, they can improve the chances of your content being considered

AI Optimization Strategies: How to make your content AI-friendly

Optimizing content to maximize the emergence of patterns learned by AI language models like ChatGPT involves making your content more understandable, relevant, and extractable for the model. To achieve this, tailor your content to align with the patterns the AI model has learned. Here's a concrete example to illustrate this process:

Suppose you want to optimize content about "The Benefits of Yoga" to be prominently featured in ChatGPT's responses.

1) Clear and concise language: Write an introduction using simple language and avoid jargon.

Example: "Yoga is an ancient practice with numerous physical and mental health benefits. It combines physical postures, breathing techniques, and meditation to promote overall well-being."

2) Logical organization: Use clear headings and subheadings to divide the content into sections.

Example:

No alt text provided for this image
Example of Logical Organization for AIO


3) Keyword usage: Incorporate relevant keywords and phrases naturally throughout the content.

Example: "Yoga helps improve flexibility, strength, and balance while also reducing stress, enhancing focus, and promoting better sleep."

4) Semantic relationships: Use synonyms and related phrases to reinforce the topic.

Example: "The advantages of practicing yoga extend beyond mere physical exercise, as it also fosters mental clarity and emotional resilience."

5) Authoritativeness: Cite credible sources, such as scientific studies or expert opinions, to support your claims.

Example: "According to a study published in the International Journal of Yoga, regular yoga practice can significantly reduce stress levels and improve overall mental health."

6) Recency and updates: Use the latest information and update your content regularly.

Example: "A recent 2023 study found that yoga may also help boost immune function, adding to the growing list of benefits."

7) Answer common questions: Address frequently asked questions related to the topic.

Example:

  • "What are the different types of yoga?"
  • "How often should I practice yoga to see benefits?"
  • "Can beginners practice yoga?"

By applying these principles and using the example as a guide, you can create content that aligns with patterns learned by ChatGPT and increase the likelihood of your content being featured in the AI model's responses. Remember that AI models are constantly evolving, so staying up-to-date with the latest developments and adjusting your strategies accordingly is crucial.

The Subtle Differences Between SEO and AIO

As AI continues to be the life of the digital marketing party, it's time to embrace AIO and optimize your content for AI-driven platforms. But what sets AIO apart from our old friend SEO? Let's break it down with some concrete examples:

SEO:

  • Focuses on optimizing content for search engine visibility
  • Targets specific keywords to rank higher in search results
  • Uses backlinks, metadata, and on-page optimization to improve rankings

Example: In our travel blog post about "Top 10 Must-Visit Destinations in 2023," we would optimize the content for search engines by using a target keyword like "best travel destinations 2023," optimizing meta tags, and obtaining backlinks from reputable travel websites.

AIO:

  • Focuses on optimizing content for AI model comprehension and reference
  • Prioritizes clear language, logical organization, and semantic relationships
  • Aims to ensure content is included in AI model training datasets and properly understood by AI platforms

Example: For the same travel blog post, AIO optimization would involve using clear language, breaking the list into subheadings, playing with synonyms, citing credible sources, and addressing common questions to ensure AI models like ChatGPT understand and reference the content accurately.

Conclusion

In a nutshell, SEO is all about pleasing search engines, while AIO is about becoming besties with AI models. As the digital landscape evolves, it's essential to strike a balance between these two strategies to stay ahead of the game and make your content shine in both search results and AI-driven platforms. So, put on your AIO hat and get ready to make your content the life of the AI party!

AIO seems similar to VSO or Voice Search Optimization for Voice Search technologies and virtual assistants for voice search results. Great content!

Maurice Ramirez, D.O., Ph.D.

Recipient of the Lifetime Achievement Award in Disaster Medicine Disabled Physician

1y

Great data for the online education division of our not-for profit as we produce evidence-based content in the All Hazards, One Health, One Nature (AHOHN) space. Simultaneously, the ability to "market" to the AI is concerning because it means organizations and individuals with and agenda have the ability to prejudice the AI in favor of their agenda-driven message. Ultimately, this article punctuates the fact that users of AI must be able to differentiate truth from disinformation. The first rule of computing is GARBAGE IN, GARBAGE OUT, but for many internet users the rule has become GARBAGE IN, GOSPEL OUT. Perhaps we should add a rule for AI Computing... CAVEAT EMPTOR... "Let the Buyer (user) Beware."

Raji Akileh, DO

Co-founder & CEO of MedEd Cloud I NVIDIA Inception | DO, Health & Wellness, Innovation, Regenerative Medicine

1y
Ben Balliett

Vice President, Head of Medical Device Sales

1y
Kevin Dent

Video games industry veteran with 20+ years of experience

1y

I see your brain turning….

To view or add a comment, sign in

Insights from the community

Others also viewed

Explore topics