Our journey to AI Search Optimization

Less than two years before this article was first published in May 2025, search still meant Google for most brands, and SEO meant optimizing for Google’s algorithm. Then generative language models arrived, able to hold conversations and give human-like answers. Today, ChatGPT has 900 million weekly users and Google’s AI Mode has more than a billion monthly users.

We didn’t want to watch that change from the sidelines. This is the story of how our team researched it, what our tests showed, and how it became a new service for our clients.

Stage I: Researching AI’s impact on search

We started by watching how people used ChatGPT and similar tools. AI was no longer just answering questions; for some searches, it was replacing search engines. We tested many types of queries and tracked new features as they appeared: category links, product suggestions, images and deep research.

That led to two questions that shaped everything that followed: are brands present in these answers, and how do they get there?

Stage II: Testing external links and content

Next, we examined which sources ChatGPT, Gemini, Claude and Copilot cited, which domains they treated as trustworthy and what influenced brand mentions. At the same time, assistants increasingly combined language models with web search: by February 2026, ChatGPT searched the web for about a third of queries.

It became clear that we were looking at a new category: AI search.

Stage III: Experimenting with content and technical optimization

We then combined SEO knowledge with the new reality of generative AI. We tested:

• which content structures AI models use most easily;
• whether length, language and format affect visibility in answers;
• how models handle schema, tables, PDFs and product descriptions;
• whether AI crawlers can see content rendered with JavaScript.

What we learned: 6 findings

Finding	What external research shows
1. AI search is mostly search	Pages ranking for fan-out sub-queries were 161% more likely to be cited in AI Overviews (Surfer)
2. Mentions outweigh links	Branded web mentions correlate with AI Overview visibility at 0.664, backlinks at 0.218 (Ahrefs)
3. JavaScript can hide content	Major AI crawlers from OpenAI, Anthropic and Meta don’t render JavaScript (Vercel)
4. Schema alone doesn’t get you cited	No meaningful citation lift after adding schema to 1,885 pages (Ahrefs)
5. Precise beats long	ChatGPT citations reward precision and ranking over length (AirOps)
6. Fresh content wins, answers vary	Cited pages are 25.7% fresher than organic results (Ahrefs); brand lists repeat in <1 in 100 runs (SparkToro)

More on these mechanisms in how AI chatbot recommendations work and 10 AI search myths.

Stage IV: Discovering a new brand visibility channel

Through this work, we saw more than a new tactic. Brands were appearing in conversations people had with AI, often before those people visited a website or filled in a contact form. AI had become an early touchpoint where users get recommendations, summaries and comparisons, and where a brand can be included or left out.

Stage V: The birth of our AI Search Optimization service

The result of this work is our AI Search Optimization service. It builds on SEO and adds four areas:

Area	What we do
Source analysis	We identify which sources AI assistants use for our client’s topics and competitors, and where the brand is missing
Content optimization	We restructure key pages so AI can quote them: direct answers, question-style headings, tables, facts and sources
Technical preparation	We make sure content is accessible: server-rendered HTML, crawlable pages, accurate metadata and product data (see structured data in AI search)
Brand presence management	We plan mentions in media, comparisons, reviews and communities, and correct outdated information (see backlinks and mentions)

Stage VI: Our internal tools

To work efficiently, we built our own scripts:

• Content scoring: evaluates content against criteria we derived from our tests and published research, such as answer clarity, coverage of sub-questions and use of facts.
• Embeddings: converts content into vectors to compare how closely passages match the questions we want to cover.
• Semantic Booster: uses semantics and knowledge graphs to strengthen the match between content and target questions (measured with cosine similarity). It’s a guide for editors, not a guarantee of citation.
• Content gap finder: uses knowledge graphs, entities and their attributes to show which topics, categories and subcategories are missing from a topic cluster (see topical authority).

What does AI Search Optimization deliver?

• presence in AI answers to questions about your products, comparisons, rankings and recommendations;
• accurate, consistent descriptions of your brand across AI tools;
• content and data that AI systems can find, read and quote;
• an early advantage in a channel many competitors haven’t started working on.

How do we measure AI search results?

Clients often ask how to measure results. AI search is less transparent than classic search, but it can be measured. We combine:

• First-party data: the Generative AI report in Search Console (impressions in AI Overviews and AI Mode), AI Performance in Bing Webmaster Tools (Copilot citations) and the AI Assistant channel in GA4 (visits and conversions);
• Prompt monitoring: a fixed set of customer questions, run repeatedly across platforms, to track mentions, citations and share of voice;
• Source and sentiment analysis: which sources AI uses, in what context the brand appears and whether the description is accurate.

Read more in metrics that matter in AI search and measuring data from generative chatbots.

What’s next?

We believe AI search is a lasting shift in digital marketing. Language models are becoming consumers’ new advisors, and brands that learn how they work will have a long-term advantage. We keep testing as models change, and we’ll keep sharing what we learn. If you’re deciding whether to start, see 5 questions to ask before investing.

FAQ

What is AI search optimization?

It’s the work of making a brand visible, recommended and accurately described in AI-generated answers from tools such as ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews. It builds on SEO.

How is it different from SEO?

SEO aims for rankings and clicks. AI search optimization also aims for mentions and citations inside answers, which depend on brand presence in other sources, quotable content and technical accessibility.

How do you get a brand cited by ChatGPT?

Rank for the main question and related sub-queries, publish clear and current answers with facts, make content readable without JavaScript, and earn mentions in trusted sources such as industry media, reviews and comparisons.

How long does it take to see results?

Content changes can appear in AI answers within weeks for tools that search the web. Building brand mentions and reputation usually takes several months.

How do you measure AI search optimization?

With Search Console’s Generative AI report, Bing Webmaster Tools’ AI Performance report, GA4’s AI Assistant channel and repeated prompt testing for mentions, citations, share of voice and sentiment.

Sources

• Surfer: Query fan-out impact on AI Overview citations (2025)
• Ahrefs: AI Overview brand visibility factors (May 2025)
• Vercel: The rise of the AI crawler (Dec 2024)
• Ahrefs: We tracked 1,885 pages adding schema (May 2026)
• Search Engine Land: ChatGPT citations reward ranking and precision over length (AirOps) (2026)
• Ahrefs: Do AI assistants prefer to cite fresh content? (2025)
• Search Engine Journal: AI recommendations change with nearly every query (SparkToro) (Jan 2026)
• Google Search Central: Search Generative AI performance reports (Jun 2026)
• Search Engine Journal: GA4 adds AI Assistant default channel group (May 2026)