TL;DR
Answer Engines (ChatGPT, Gemini, Perplexity) already answer queries that previously generated clicks to your site. According to Avinash Kaushik's model, your company could lose up to 30% of organic traffic by 2026. Informational queries are the most affected (up to 60% loss). But there are concrete actions that recover between 10% and 15%: schemas, authority, and comparative content. Below we include an interactive calculator for you to estimate your case with real data.
AI no longer complements search. It replaces it.
Until recently, searching on Google was an act of exploration: you typed keywords, checked 3 or 4 results, went back, adjusted the query. Today the dynamic is different. A user types a long, context-rich question into ChatGPT and receives a complete answer without leaving the interface. No click. No visit. No session in your GA4.
As Cristóbal and Catalina from the consultancy D2B explained in a recent micro-training for the SearchBrand community, what used to be a librarian providing you with books to research, has now become a post-graduate trained assistant giving you the final answer.
The data supports this change. According to information presented by D2B from their client base in Latin America (e-commerce, finance, education), traffic from LLMs already accounts for between 0.02% and 0.03% of total sessions, and is growing at a rate of 5x per year. ChatGPT accounts for 95% of that referred traffic from language models.
Which industries lose the most traffic (and why)
Not all queries are equal when it comes to AI. Informational questions, those starting with "what is," "how it works," or "what's the difference between," are those that LLMs resolve most easily. And they are, therefore, the ones that stop generating visits.
According to D2B data, the most exposed sectors are health (symptoms, medications, conditions), education (admission requirements, program comparisons), and news. In these industries, the LLM's answer is enough for the user not to click.
E-commerce and financial services are in an intermediate zone: transactional queries still require the user to visit a site to buy or contract, but the preliminary research phases are already resolved within the chatbot.
Avinash Kaushik's Loss Model, Step by Step
Avinash Kaushik, the digital analytics expert who created the Occam's Razor blog and worked for years at Google, published a loss and recovery forecasting model for the Answer Engines era at the end of 2025. Cristóbal and Catalina from D2B adapted it for the Latin American context and presented it in their micro-training.
The logic is this: not all your organic traffic is equally threatened. It depends on the search intent behind each keyword. The model classifies queries into four types and assigns a potential loss percentage to each:
| Query Type | Example | Risk | Potential Loss |
|---|---|---|---|
| Informational | "what is a checking account" | High | 60% |
| Commercial | "best bank for SMEs in Chile" | Medium-High | 35% |
| Non-branded transactional | "open checking account online" | Moderate | 15% |
| Branded / Navigational | "Tenpo digital account" | Low | 5% |
Why do informational queries lose more? Because AI can answer them completely. There's no reason to click. Why do branded queries lose little? Because if you search for "Tenpo" or "Falabella", the intention is to go to that site, and the LLM knows it.
How to Perform the Calculation with Your Own Data
You don't need a data science team. You need Search Console, GA4, and a spreadsheet. Here are the steps Catalina from D2B presented in the micro-training:
- Export your keywords from Search Console (last 90 days or 6 months). You need the complete list with clicks for each query.
- Classify each keyword by intent: informational, commercial, non-branded transactional, or branded. This step is manual but important: the quality of the calculation depends on this classification.
- Sum the clicks for each category and calculate the % share of each over the total.
- Obtain organic sessions and organic revenue from GA4 for the same period. Use the sessions report (not users) filtered by the organic channel.
- Calculate SEO value per visit: total organic revenue divided by total organic sessions.
- Redistribute GA4 sessions according to the percentages of each category (as calculated in step 3).
- Apply the loss percentages from the table above to each category. This gives you the projected loss in sessions and revenue.
A key insight Cristóbal mentioned: the allocation of Search Console clicks to GA4 sessions is not perfect. They are different tools that measure slightly different things. But the percentage distribution by search intent is the best available proxy.
AI Traffic Loss Calculator
Enter your real data from Search Console and GA4. The calculator applies Kaushik's model, adapted by D2B, and shows you how much you could lose in sessions and revenue.
📊 AI Impact on SEO Traffic Calculator
Based on Avinash Kaushik's loss model, adapted for LATAM
Clicks by Category (Search Console)| Category | Est. Sessions | Loss % | Sessions Lost |
|---|
* Recovery percentages based on Avinash Kaushik's model. Actual results depend on speed and implementation.
Watch the complete micro-training
Cristóbal and Catalina from D2B explain the model step-by-step, with examples and the live template. This article is based on their presentation.
How to Recover Some of That Traffic (the model's 6+6 actions)
Losing traffic doesn't mean standing idly by. Kaushik's same model proposes 6 recovery actions and 6 growth tactics, each with an estimated gain percentage. Catalina summarized them as follows in the micro-training:
6 direct recovery actions
Adapt content to function as an AI answer (questions in H2, direct answers in the first paragraph). Implement JSON-LD schemas (product, FAQ, HowTo, organization, reviews), with an estimated 4% recovery. Build topical authority with author biographies and structured data, for a 3% recovery. Optimize branded queries to control rich results. Create comparative content where your brand stands out against the competition. And generate practical guides with use cases and differentiating benefits.
6 additional growth tactics
Group content that solves the same problem (topic clusters). Create content for specific use cases. Scale authority with third-party content and link-building strategies. Incorporate video and audio about products. Integrate reviews as first-party content (not Facebook or Google embeds). And prepare the site for AI-driven commerce scenarios (structured feeds, complete product data).
For PPC teams, Cristóbal mentioned estimated losses of 15-20% in paid traffic, with recommendations to evolve copywriting towards useful answers (not just sales), migrate to Performance Max where conversion rates and ROAS are considerably higher than traditional search, and feed platforms with first-party data to improve targeting.
Recovering traffic starts by measuring where you are today
Calculating potential loss is the first step. The second is to know what AIs are saying about your brand today. Do you appear when someone asks ChatGPT about your industry? What does Gemini say about your competitors?
Tools like SearchBrand.ai allow you to monitor these questions in real-time: queries to multiple LLMs, analysis of whether your brand appears or not, what sources they cite, and a comparable visibility score over time. For brands in LATAM that are just evaluating their position in this new channel, this initial diagnosis is the starting point before implementing any recovery action.
Sources and Methodology
This article is based on the SearchBrand community's micro-training, presented by Cristóbal and Catalina from D2B, a digital marketing consultancy with 7 years of experience in LATAM. The loss and recovery percentages follow the model published by Avinash Kaushik. Data on referred traffic from LLMs comes from D2B's client base (e-commerce, finance, education). Market figures were verified with external sources.
- SearchBrand Micro-training: How to quantify the impact of AI (Full Video)
- Loss Recovery Growth Model: Answer Engine Optimization (AEO), Avinash Kaushik, Occam's Razor (Dec. 2025)
- AI Age Marketing: Bye SEO, Hello AEO!, Avinash Kaushik, Occam's Razor (July 2025)
- Gartner: projection of 25% of search volume migrating to AI chatbots by 2026
Frequently Asked Questions
How much organic traffic will be lost due to artificial intelligence by 2026?
According to Avinash Kaushik's model, companies could lose up to 30% of their organic traffic during 2026. Informational queries are the most affected with up to a 60% potential loss, while branded queries have a low risk of 5%.
How do I classify my keywords by search intent?
Informational queries answer questions ("what is", "how it works"). Commercial queries compare options ("best X for Y"). Non-branded transactional queries seek action without mentioning a brand ("buy", "hire"). Branded queries include your brand name or product. Export your queries from Search Console and classify them manually or with AI assistance.
Can traffic lost due to AI be recovered?
Partially. Implementing structured schemas, building topical authority, and creating comparative content can recover between 10% and 15% of lost traffic. The key is to get Answer Engines to cite you as a source and link to your site.
What data do I need to use the loss calculator?
You need clicks by intent category from Search Console (informational, commercial, transactional, branded), total organic sessions from GA4, and revenue attributed to the organic channel. All for the same time period.
What is Avinash Kaushik's loss model?
It is a framework published in the Occam's Razor newsletter that classifies organic traffic by intent and assigns potential loss percentages to each type. It also proposes recovery tactics with estimated gain percentages, allowing companies to build a complete forecast of AI's impact on their business.