marketingartificial intelligence
AI Search Optimization Can't Wait: Why Marketers Must Adapt Now
Global State of Ecommerce 2025 report found that 11.4% of retail traffic from ChatGPT converted to sales, compared with 9.3% for paid search and 5.3% for organic search.
While good SEO contributes to AI referrals, the winners are not always the same. For example, Ally Bank gets more referral traffic from ChatGPT than traditional banks. Why? Because ChatGPT prioritizes well-reviewed, user-friendly products, and Ally Bank consistently performs well in organic sources like Reddit and review sites that ChatGPT favors. So far, there is no ChatGPT advertising that would allow brands to buy a place at the top of the banking recommendations.
Think conversation, not conversion
While high conversion rates for AI are interesting, we can’t just focus on clicks that drive traffic with GenAI the same way we do with search. Traditionally, users who entered search queries were looking for links to click on. That’s changed as AI summaries become common on search engine results pages, driving up zero click behavior. It’s changed even more with GenAI, where links are presented as footnotes to a chatbot’s answers.
That makes it doubly important to understand the types of conversations consumers are having with AI engines. Through repeated prompts, answers, and follow up questions, they are learning about your brand and products through pre-digested AI summaries of your content in combination with independent news, reviews, and social posts that either support or undermine your marketing efforts.
That’s how the AI decides whether your website, blog post, or product page is worth linking to – and how the consumer decides if the link is worth clicking on. Instead of clicking, they might research your brand further through conventional search – about 95% of ChatGPT users also continue to use Google – but only if the AI answer motivated them to learn more. Getting traffic as a result of these interactions is important but so is the opinion consumers form about your brand without even visiting your website or your app.
In other words, conversation comes before conversion.
Just as the exact rules of SEO have to be rewritten every few months to account for algorithmic changes, GenAI technology will continue to change. But SEO experts have established some foundational principles over the years, and we need to do the same for AI.
Here is what we’ve learned from our own explorations and consultations with industry experts.
AI prompts are highly personal and less generic than search keywords
For decades, we’ve been building SEO tools and strategies around identifying common keywords and optimizing content to match. Users may misspell words or use slightly different phrasing, but there’s a reasonable degree of convergence around keywords like “best skincare routine.”
AI prompts tend to be much more specific. Instead of a couple of skincare keywords, a user will often provide their age and details about other products they have tried and what has and hasn’t worked for them. Rather than asking for links, they ask for detailed recommendations and a plan of action.
If you can get access to sample prompts, they can be invaluable for market research. On the other hand, it’s unlikely that thousands of people will enter the exact same prompt with the exact same details. Rather than thinking in terms of search query keywords, we need to identify common themes for prompts. In the example shown below, we’re exploring answers to questions about vacuum cleaner models, their positive and negative sentiment, and specific product attributes different brands win on.
Early theories of AI optimization
For guidance on how optimizing for GenAI is different, allow me to share the insights from experts who participated in a recent Similarweb Search Summit in London: Kevin Indig, who has been providing regular updates on the impact of AI on search through his Growth Memo newsletter and Aleyda Solís, founder of SEO consultancy Orainti, who provided an AI search optimization checklist. Here is a mashup of their recommendations:
- Discover the right social and discussion platforms for the focus where your brand needs to win influence, prioritizing those most frequently referenced in ChatGPT answers and Google search results for key topics.
- Simplify and structure your branded content to make it easily digestible by both people and machines. For better placement in AI summaries, Solis recommends “chunking” content into focused semantic units that can stand on their own, independent of a larger document.
- Aim for recognition as a trusted, clearly differentiated brand. Cultivate proof points (success stories, benchmarks, certifications), experts advocates, validation by partners, analysts, reviews and press, and consistent visibility everywhere that potential customers are forming opinions and making decisions.
- Monitor your brand visibility, sentiment, referrals from AI to identify opportunities to improve mentions that matter versus competitors.
Example: Winning in skin care
An example of a brand with a winning share of AI referral traffic is The Ordinary in skin care. Based on U.S. web traffic, March to August 2025, theordinary.com captured a 57.56% share versus several close competitors.
An analysis of typical prompts related to this traffic shows that consumers are often either asking for the brand by name or asking questions related to the characteristics of its best known products (like “Can I use Niacinamide 10% + Zinc 1% serum with vitamin C products?”). Likely that’s because The Ordinary includes specific ingredients in their product names, where competing brands like Cerave are more likely to use terms like “foaming cleanser.”
As a result, The Ordinary has gotten out the message that these are desirable skincare ingredients, and consumers are asking detailed questions about them.
ChatGPT has gotten the message not only from the brand website but from external news and review sources. So has Google search, which includes an AI Overview summary citing The Ordinary’s content and places theordinary.com at the top of its search engine results, followed by links to a Healthline article and forum posts on Quora and Reddit.
The lesson: Paying attention to what works in GenAI visibility and how that is different from SEO is important. At the same time, improving your content and your external mentions to boost your GenAI standing is also likely to boost your search ranking – particularly as GenAI features are more tightly coupled with search.
Yes, it’s true we’re very early in the process of understanding how to optimize for GenAI. No, it’s not too early to get started.
Don’t wait.
