The Guide to AI Optimisation (AIO)

3 March, 2025

Let’s get to grips with a new era of online search.

Optimizing the output of AI is almost an entirely new concept, so stay with us as we try to navigate and find the best way to explain this! Let’s start with understanding the basics- what is AI optimization and why on earth should brands be thinking about it.

It’s easy to fall into the trap of believing that AI, with all is infinite wisdom ahem, will know automatically know what your brand is, who you are and what you do. But would it surprise you that even for some of the biggest brands on earth AI overviews are getting this information wrong?

The problem is that AI wants to serve you, and will even lie to do this. It perceives your brand as a single entity and summons all the knowledge it can muster by crawling your site or referring to more static training data. If you have not put in the work to make sure this data is machine readable and completely correct then AI could be hurting your brand before people have even accessed your website.

How can AIO improve AI’s understanding of your brand?

AIO is essentially the processes you need to go through to target AI’s fundamental understanding of your brand on a technical level. This framework, which should technically be called AIKO given its focus on optimising the knowledge LLMs have about you, will be referred to as AIO throughout this post for simplicity. It uses targeted strategies and techniques to fine-tune how AI models understand and represent a brand.

In marketing, a brand goes beyond just a name or logo. It embodies the unique identity, values, reputation, and emotional connection that a business builds with its audience. This identity influences user intent and shapes how customers perceive and interact with the business.

In the realm of AI models and Natural Language Processing however, a brand is treated as an “entity”: a distinct unit of meaning that the system recognizes and associates with specific attributes and context.

By optimizing the AI’s knowledge about these entities, AIO ensures that the system can accurately identify, retrieve, and present information about a business. This leads to more reliable, error-free responses and enhances the visibility of the brand in AI-driven searches and interactions.

In essence, AIO bridges the gap between the rich, multifaceted concept of a brand in the marketing world and the technical, data-driven understanding required by modern AI systems.

Why AIO Matters

AIO is to AI-driven search what SEO is to Google Search. Instead of ranking higher, AIO ensures AI models understand and represent your brand correctly. If AI misconstrues your brand, potential customers, investors, and media may receive misleading information.

It may not seem like a big threat now, but as the next generation grows up relying more and more on AI it could become one of the most impactful shifts you make in your online brand alignment strategy.

How to Implement AIO

AIO is a three-step process:

1. Diagnose knowledge issues that Large Language Models (LLMs) may have about a brand.
2. Identify the cause of the issue.
3. Implement solutions to fix and optimize the AI’s knowledge about the brand.

Note: If you haven’t read “Why AI Training Data is the New Position Zero,” we recommend doing so. It explains key concepts such as the two types of AI search methodologies: Grounded Search ( Live Lookup ) and Standard Search ( Internal Knowledge Retrieval _).

Step 1: Diagnosing AI Knowledge Errors

AI knowledge errors typically fall into four categories. These errors can be detected by querying LLM training data multiple times, ensuring that responses come from training data only rather than live search data. Tools like Waikay simplify this process.

1.a. Hallucinations

Hallucinations occur when AI generates completely incorrect or fabricated information. This can manifest in several ways:

• Incorrect or fabricated source links: When AI relies only on training data (approximately 50% of cases), it may generate non-existent sources.
• False product features or misinformation: AI may introduce features that do not exist or make incorrect assumptions about company acquisitions or ownership structures.

Example: For InLinks (a provider of Waikay technology), AI made the following hallucination errors:

• Invented a non-existent SERP tracking feature for InLinks
• Incorrectly stated that SEMrush acquired a core component of InLinks and rebranded it as SEMrush’s Content Writer Assistant.

Even real-time search AI models (e.g., Gemini Grounded) have replicated these hallucinations:

1.b. Knowledge Disappearance

AI may fail to acknowledge key products or services from a brand. A notable example involves a major French retailer of construction materials:

• The retailer sells 130,000 electrical products and 30,000 plumbing fixtures.
• However, ChatGPT never mentioned electrical products, even though they make up the largest portion of their catalogue.

The exact same thing occurred for a major SEO tool (one from the top 5), for which one of its prominent tool doesn’t simply exist in Gemini training data.

1.c. Knowledge Contradictions

This occurs when AI provides conflicting information on the same topic.

Example:

• InLinks is a pioneer in entity-based SEO, with extensive published resources and a book written by its CEO.
• Despite this, ChatGPT provided contradictory definitions of Entity SEO, confusing it with alternative concepts.

1.d. Pure Ignorance

In some cases, AI simply has no knowledge of a brand, its products, or services.

This happens more frequently than expected. For example, Waikay’s AI Knowledge Score analysis found that while ChatGPT recognized major SEO tools well, Gemini Training Data (TD) lacked any awareness of them.

0% Gemini Knowledge Score for these 3 well known (sorry, hidden) SEO tools.

Step 2: Identifying Causes of Knowledge Issues

The next step involves analyzing the brand’s discourse and semantics to determine where knowledge gaps originate. This process is closer to a literary and linguistic analysis rather than just SEO.

2.a. Discourse Discontinuity

A product or feature that has been discontinued and then relaunched can lead to knowledge gaps.

Example:

• If a relaunch isn’t properly documented across web content and external sources, the LLM may assume that the product no longer exists.

2.b. Discourse Contradictions

Conflicting brand mentions can confuse AI models.

Example:

• A multimillion-dollar SaaS company was mistaken for a car brand in different AI models.
• ChatGPT Training Data provided the correct response, but Claude (Anthropic) and Gemini incorrectly associated it with Bentley Motors.
• Bentley Motors never owned the Bentley.com domain, but widespread online mentions of the car brand influenced AI responses.

And here’s what Gemini says using its training data:

“Based on my training data, I know that bentley.com is the official website of Bentley Motors Limited, a British luxury car manufacturer. I know this because the domain name itself strongly suggests this affiliation, and this fact is widely known and documented across numerous reliable sources on the internet (though I cannot directly cite them here as I don’t access live websites or specific articles).”

Scary, no?

2.c. Discourse Shortcomings

This happens when a brand restricts AI’s ability to access its website, preventing accurate knowledge acquisition.

Example:

• Microsoft blocks AI from scraping its website, which limits AI’s ability to understand real-time product offerings.
• While this isn’t a major issue for Microsoft due to widespread press coverage, smaller brands relying on niche content could suffer.

2.d. Wrong Semantic Positioning

AI models may misinterpret category structures and fail to retrieve relevant content.

Example:

• The French retailer of construction materials listed electrical products under “Electrical Equipment” instead of “Electricity.”
• Since “Electrical Equipment” is a subcategory of electricity, AI failed to recognize it as a main product category.

Step 3: Fixing LLM Knowledge Issues

3.a. Adapting Online Content

• Ensure SEO-friendly and AI-friendly content coexist by optimizing terminology.
• AI understands structured content best, so optimize key landing pages with schema markup.

3.b. Expanding Online Content

• AI prioritizes expert-level, long-form content over promotional material.
• Example: Instead of writing “Why our tool is the best SEO software,” publish “The Evolution of SEO Tools & Their Impact” while subtly referencing your brand.
• Avoid low-quality AI-generated content, as it dilutes credibility.

3.c. Influencing Online Content

• Increasing brand mentions in authoritative publications ensures AI has access to trusted external references.
• Actively engage with Wikipedia editors to maintain updated, verified brand information.

3.d Securing Your Data Against AI Misinterpretations

Some brands unintentionally restrict AI models, causing knowledge gaps.

• Allow AI Bots Access – don’t block AI web crawlers (e.g., ChatGPTBot, BingBot) in robots.txt. Instead, whitelist critical sections of your site for AI access.
• Publish AI-Friendly FAQs & Knowledge Hubs – create AI-optimized FAQ pages to answer common queries about your company.
• Monitor AI-Generated Summaries – ensure Google’s Knowledge Graph and other AI-powered summaries display accurate information about your brand.

4. Monitoring AI Knowledge Over Time

• Conduct continuous AI knowledge audits.
• Use AI Knowledge Monitoring tools like Waikay to track brand presence in AI-generated responses, for the most used LLMs. Go beyond the simple brand level verification and conduct AI Knowledge Audits about your most important topics
• Submit AI feedback to OpenAI, Google, and Microsoft to report errors.

Recap: Ensure AI Understands

• Identify AI knowledge gaps via structured queries across multiple AI models.
• Maintain consistent brand discourse across all digital assets.
• Optimize Wikipedia, Wikidata, and high-authority sources.
• Use structured data and FAQ schema to improve AI information retrieval.
• Regularly audit AI responses and report errors.

Final Thoughts

AI Optimization is as crucial as SEO in the AI-driven search era. A brand that fails to secure its knowledge in AI models risks misinformation or omission. As the internet progresses and the world becomes more reliant on AI, being misunderstood could be one of the most detrimental things for your brand.

By taking proactive AIO measures, businesses can ensure accurate representation in AI-generated search results and gain a competitive advantage in the AI-driven digital world.

Written by Fred Laurent, reviewed by Genie Jones