Peec AI tracks how LLMs pick brands in chunks
AI search has become a channel brands have to manage, and the unit of competition is smaller than a page. Models split a document into chunks, score each chunk and cite only the ones that stand alone, so a paragraph that leans on its context carries no information. Peec AI runs customer prompts against ChatGPT and Gemini daily, records which brands and sources appear, and tells clients which third-party pages they are missing. Landwehr argues the durable edge is research into how the models behave, which is why a slice of the company does nothing else.
Takeaways
- LLMs score chunks of a page, not documents. When a model downloads a page it splits the text into chunks, scores each one and uses only some. Landwehr says content should be rewritten so chunks are self-contained and verifiable, since a paragraph that depends on its context carries no information.
- Peec AI simulates daily prompts to track brands. It agrees prompts with each customer, types them into ChatGPT and gemini.google.com every day, then records which brands and sources appear. It downloads those sources and flags third-party gaps such as a missing G2 page or a missing Forbes interview.
- LLMs ground answers in consensus across sources. Models look for agreement across the web, so a brand absent from the pages they read goes unmentioned. Peec AI checks LinkedIn, Instagram, Reddit and third-party sites where competitors are cited but the customer is not.
- A research team sets Peec AI apart. Landwehr says 3% of the company works only on how AI search works, on top of the data science team. That is why Peec AI can add a new Gemini model the day it appears and ship news about AI search changes.
- AI search lead share is climbing fast. B2B SaaS companies publicly reported getting 15% of new leads from AI search, then 20% a few months later. Agencies were the first customers and the base is now roughly fifty-fifty agencies and in-house teams, with travel and e-commerce growing on ChatGPT's shopping feature.
- Peec AI hit $7.5m ARR and 100 employees. The company has close to 4,000 customers and grew from five to seven people two years ago to more than 100 employees. Larger clients pay six figures a year through sales, while self-serve signup and product-led growth carry the rest.
Peec AI's plans to the end of 2026
- Pull in far more data, including sources like Google Search Console.
- Big agentic improvements in the product.
- More users running their AEO and SEO prompts in Peec AI's agent instead of Claude, because the agent carries more context.
- More integrations.