Influence of SEO on AI Overviews
At the beginning of the week, I came across a post on Reddit titled “I prefer authentic search results.” It had over 62,000 upvotes and hundreds of responses, all about the dislike of AI Overviews.
Phrases like “information is wrong,” “contradiction,” “no fact checking,” and “I want sources, not pasted together answers” were repeated repeatedly.
For me, such feedback is wonderful; we learn through feedback, specifically negative feedback.
While positive feedback is wonderful, it doesn’t tell us anything about how to improve.
What does this have to do with SEO and AI Overviews? I’m glad you asked.
Have you ever used a GPT like OpenAI’s GPT-4o, Gemini, or Llama? Not the Perplexity or ChatGPT type, but the underlying model?
Services like ChatGPT.com, Perplexity, and Google’s AI Overviews are wrappers that add capabilities such as internet search, and recommendation, the bits we consumers want.
The model itself is a form of AI called Natural Language Processing (NLP) that understands only the structure and semantics of language.
They do not create indexes or ranks of web pages, they embody complex language patterns and relationships designed to understand a language.
Let's say you ask a GPT for a product recommendation and to cite its sources.
Given your prompt/query, it would generate a response including a URL. The URL will most likely be nonsense, a complete fabrication. The model isn’t lying, it’s just predicting the next word in a sequence given all the words that came before it.
It’s not that the fabricated URL is junk; when the model was trained, the domain would have been associated with a body of text that the training process used. The domain is valuable as a piece of metadata as it offers a road sign to where the training process sourced learning material from.
Regardless of how and where an AI Overview or response is presented, the GPT's role is to use its understanding of the language, interpret the context (the input query), and generate the most likely output.
This is an important point. Given an input, we ask the GPT to infer the best output.
SO ...
When it comes to AI Overviews, we need to ensure that we provide the best possible input data to ensure that whoever receives the output is pointed in the right direction—specifically, the brand, product, or service that we’re promoting.
And this is where SEO comes into play. It is also why AI Overviews change.
Like ChatGPT and Perplexity, the software behind AI Overviews passes information pulled from the search index to the model. The model takes this information and, using its understanding of language, returns a response.
Unless told otherwise, the model will always be unaware of search indexes, ranking, the relevance between one topic and another, one interpretation, and another. All it knows are the semantic rules governing a language.
How do you influence an AI overview?
· Make sure that you have a commanding influence over the information passed to the model.
· That your SEO is the best that it can be.
· That you are seen as authoritative and trustworthy across many sources, not just one.
· You understand what semantic rules the model deems important.
And this is what we’re working on at GetHeard 😊.
