How to calculate how much organic traffic you'll lose to AI (and how much you can win back)

On January 8, 2026, I hosted a SearchBrand micro-training with Cata and Cristóbal from D2B, a Latin American consultancy. Eleven minutes in, they dropped a Google Sheets template into the chat. Cata typed click totals for four query groups (200 informational, 100 commercial, 600 transactional, 50 brand), 20,000 organic sessions and 280,000 in revenue. The sheet immediately showed how many sessions and how much revenue each intent group stood to lose to answer engines.

This article replaces the previous version of this guide. The loss rates now match what the template actually showed on screen, and the ones it didn't show are labeled as assumptions. The calculator reads your Search Console CSV directly and includes a module to estimate how much you can recover.

The webinar is in Spanish. Timestamps below point to the relevant moments.

What D2B sees in traffic from ChatGPT and other LLMs

Cristóbal co-founded D2B, a firm that has spent 7 years doing SEO and CRO work, plus analytics and performance, for companies across Latin America. He opened with what they saw across clients as of December 2025, a sample heavy on ecommerce with some finance and education clients: sessions from LLMs were still a very small slice of total traffic, and probably smaller on informational sites (3:30).

ChatGPT accounted for almost all of it. Gemini came second, with Copilot and Perplexity at very low shares. His advice for teams with an endless backlog: prioritize ChatGPT and keep an eye on ads inside ChatGPT, which he expected this year. Check where that stands in your market before you plan around it.

I'm not repeating Cristóbal's exact percentages. The auto-generated transcript makes them ambiguous and D2B didn't share its sample size. Your own GA4 will give you a better number anyway.

Which industries lose the most to zero-click

According to Cristóbal (5:55), exposure depends on how easily the answer can be summarized:

Site or query type	Exposure per D2B
Informational content and news sites	High
Queries whose answer fits in a paragraph	High
Health and education, given how much they rely on informational search	High
Transactional and financial	Lower: the shift "isn't as drastic"

D2B gave no per-industry percentages, so there are none here. Use the table to pick a calculator preset.

From librarian to research assistant

Cristóbal used an analogy worth stealing for your next internal deck. Google acts like a librarian: it hands you references and you do the browsing, the back button, the next click. An answer engine is closer to "an assistant with graduate-level research skills" (6:40). His example was a long prompt asking for sneakers in a specific size, without certain features, avoiding certain brands. What comes back reads like a final recommendation.

His summary of the traffic effect: where you used to visit somewhere between three and eight sites before deciding, a couple now do the job. The model below tries to put a number on that gap.

Before you calculate: the data you need

The calculator needs three inputs, and all three must cover exactly the same date range. Cata and Cristóbal hammered on this between 14:30 and 18:00.

From Search Console: under Performance → Search results, pick the date range, open the Queries tab and export to CSV. The file can come in English or Spanish, and the decimal separator doesn't matter. The UI export tops out at 1,000 rows and leaves out anonymized queries, so the sum of clicks in the file rarely matches the chart total. The calculator shows you that gap.

From GA4, Organic Search sessions. Sessions, not users. Cristóbal's reasoning: a Google click maps much more closely to a session than to a user, who may come back several times. In practice that's mostly google / organic, which he called "probably the 80/20," plus Bing and other engines.

From GA4, revenue for the same channel and range. That gives the calculator your value per visit.

Which window? D2B uses 90 days or 6 months depending on the client. For seasonal businesses Cristóbal recommends a full year, and gave the example of a business that peaks in February and March: a 90-day window from the off-season skews everything. If your last 12 months of tracking are broken, use whatever range is clean. I default to 12 months whenever the data exists, because the number you'll present is an annual one.

Calculator: upload your Search Console CSV and edit the assumptions

The calculator opens with the webinar demo preloaded. Switch to "Upload Search Console CSV," type your brand and its variants (with and without spaces, common misspellings, your own product names) and review the classification table: any query can be moved to another group in one click. Presets are explained in the next section, and every rate can be edited.

Once you see where you lose the most, the next question is whether answer engines already cite you for those queries. SearchBrand measures that: create an account with 10 free credits, no card required.

The model in five formulas

Cata credited the model's structure to Avinash Kaushik: four intent groups, each with its own loss potential (11:20). D2B's template and this calculator do the same math:

share_i        = clicks_i / total clicks
sessions_i     = share_i × GA4 organic sessions
value_per_visit = organic revenue / organic sessions
lost_i         = sessions_i × loss rate_i
lost_rev_i     = lost_i × value_per_visit

Step two is the weakest link, and D2B said so plainly. You can't join Search Console and GA4 query by query, so the model assumes each group's click share is also its session share. When someone in the audience asked about it, Cristóbal answered: "We have no way to join them perfectly" (31:00).

Value per visit is the classic visit value: channel revenue divided by channel sessions. With 280,000 in revenue and 20,000 sessions, that's 14 per visit.

Where each rate comes from (and why we changed this post's numbers)

The previous version of this article used 60% for informational, 35% commercial, 15% transactional and 5% brand, and credited all four to Kaushik. Rewatching the session, D2B's template applies 45% to the informational group: 4,210 informational sessions yield 1,895 lost sessions (27:40). The 35% commercial rate does match. The transactional and brand rates never appear on screen.

We also couldn't verify that any of these figures come verbatim from Kaushik. So each preset states its source:

Intent	Conservative	Base	Aggressive
Informational	32%	45% (D2B template)	60% (previous post)
Commercial	25%	35% (D2B template)	45%
Non-brand transactional	10%	15% (previous post)	20%
Brand	3%	5% (previous post)	7%

Unlabeled values are SearchBrand assumptions: base × 0.7 or × 1.3, rounded. Cata was explicit that the rate is meant to move. At D2B they adjust it by client and industry, and they also look at each group's CTR (12:40). The calculator has an optional adjustment for that: if a group's CTR falls below your threshold (2% by default), it adds 10 points to that group's loss rate, on the theory that AI answers are already absorbing those clicks. It ships switched off because the right threshold depends on your category.

Step-by-step example

The webinar demo, reproduced

These are the numbers Cata entered live. On the base preset, the calculator matches the template for the two groups shown on screen:

Intent	Clicks	Sessions	Loss rate	Lost sessions	Lost revenue
Informational	200	4,210	45%	1,895	26,526
Commercial	100	2,105	35%	737	10,316
Transactional	600	12,632	15%	1,895	26,526
Brand	50	1,053	5%	53	737
Total	950	20,000	22.9%	4,579	64,105

The transactional and brand rows use our assumptions, so the total isn't D2B's. In clicks, the loss comes to 218 out of 950.

Notice that the transactional group loses as many sessions as the informational one at a third of the rate. It holds 63% of clicks. In ecommerce that group tends to be the biggest, which is why a couple of points on its rate move the total more than 15 points on informational.

A sporting goods store with 12 months of data (illustrative)

This case isn't from the webinar. We built it with round numbers to walk through the full flow.

1. Export 12 months of queries from Search Console. Say 1,000 rows and 18,000 clicks.
2. Enter your brand terms: the store name, the no-space version, the most common typo. Result: 3,600 brand clicks.
3. Let the rules sort the rest. "price," "buy," "deal" and "shipping" go to transactional; "best," "vs" and "reviews" to commercial; everything else stays informational. Result: 7,200 informational, 3,600 commercial and 3,600 transactional.
4. Review the reconciliation check and the table. "how to choose trail running shoes" belongs in informational, but "men's trail shoes" has no modifier and lands there too. Move it to transactional if that search buys in your category.
5. In GA4, open Traffic acquisition, filter to Organic Search and note sessions and revenue for the same 12 months: 60,000 sessions and $120,000, or $2 per visit.
6. On the base preset, gross loss is 17,400 sessions (29%) and $34,800. Informational queries account for 10,800 of those sessions.
7. Under recovery, check schema (4%) and authority (3%). Applied to 60,000 sessions that's 4,200 sessions and $8,400 over 90 to 120 days.
8. Net loss lands at 13,200 sessions (22%) and $26,400. Between the conservative and aggressive presets the range runs from 8,040 to 18,840 sessions, or $16,080 to $37,680.

Net loss with its range is the figure you bring to a budget meeting.

Winning traffic back: six actions and what each is worth

Midway through, I asked Cristóbal whether the model accounts for brand visibility you gain even as clicks drop (24:35). He said yes: just as some query types carry losses, some actions carry gains. Then came the caveat that estimating SEO improvements "has always been super hard," and that these estimates get corrected over time.

Recovery comes from being cited. Cata framed the goal as content clear and trustworthy enough that answer engines choose to cite it and link to it for more detail (32:50).

The six actions she walked through, credited to Kaushik's methodology:

1. Make content work as an answer: question-shaped H2s, a direct answer in the first paragraph and, for ecommerce, answered questions on product and category pages.
2. Implement schema (Product, Organization, FAQ, HowTo, Review, AggregateRating).
3. Build authority, starting with author bios.
4. Keep structured feeds up to date: Merchant Center and the product feed for shopping inside ChatGPT that Cata mentioned. OpenAI's program changed during 2026, so check the current terms before investing.
5. Own your branded queries and their rich results, so the source about your brand is you.
6. Publish comparisons and practical guides. If you can't name competitors, cover use cases and differentiators AI can use to compare.

Only two come with a number. Per Cata, Kaushik estimates that implementing schema correctly recovers 4% and building authority 3%, over 90 to 120 days (36:50). She added that these are baseline figures from Kaushik's own projects and shift as you measure.

The slide left it unclear what that percentage applies to, so the calculator spells it out: by default, total organic sessions (4% of 20,000 is 800 sessions). For a more cautious read, switch it to "lost sessions." The other four actions start at 0% until you have your own data.

A note on item 2: since 2023 Google shows FAQ rich results only for government and health sites, and it retired HowTo rich results. The markup still helps machines parse the page, but don't count on the rich result. To decide where to start, run an AEO and GEO audit.

More growth tactics

Cata added a second list, also credited to Kaushik and without percentages:

• Cluster content that solves the same problem.
• Write by use case.
• Scale authority with third-party content.
• Link building.
• Product video and audio.
• First-party reviews collected in your own CMS instead of Google or Facebook widgets.
• Get the site ready for AI agent commerce.

What about Google Ads?

Cristóbal said the same model runs on Search and Performance Max, and that paid traffic losses are also expected, though D2B wasn't seeing them in client data yet (40:20). The calculator has an optional module for that, with a rate you set.

His four PPC recommendations: ads that answer instead of just selling (his counterexample was "Fly to Brazil for 1,000, all-inclusive"), shifting budget from Search to Performance Max, using content URLs as audience signals, and feeding campaigns first-party data from the CRM and email lists. What he said about PMax beating Search on conversion and ROAS is agency experience, not a benchmark.

What we don't recommend

• Presenting a single number. The model rests on proportional attribution and on rates nobody has validated for your category. Always show the range.
• Using users instead of sessions, or pairing one Search Console range with a different GA4 range. Either mistake distorts the loss more than any preset.
• Applying 60% to informational queries by default for a store whose traffic is mostly transactional. That value makes sense as a ceiling for a news or health site.
• Filling all six recovery actions with made-up rates until net loss hits zero. No data, leave it at 0%.
• Running the numbers without checking the actual trend. If your organic traffic over the last 12 months isn't falling, say so: Cristóbal noted some industries aren't seeing the effect yet.

Taking the number to leadership

Cristóbal's close was about the internal conversation more than the math (45:00): share and debate these estimates with your team before you take them upstairs, explain to managers and the C-suite why the loss will happen using data, and don't show up with only the loss. The calculator's downloadable report includes the assumptions and the date range so your team can challenge them first. In the meeting I'd use three rows: gross loss, expected recovery at 90 to 120 days, and net loss, each with its range.

Clicks lost tell you where it hurts. The next step is finding out whether AI answers cite your competitors on those queries. SearchBrand tracks that with your own prompts: 10 free credits, no card required.

Sources

• Webinar "From SEO to AEO: winning in the age of answer engines (practical model)," SearchBrand with D2B, January 8, 2026, in Spanish: https://www.youtube.com/watch?v=BkAj-pDQFL0
• Avinash Kaushik's methodology as presented by Cata and Cristóbal in the webinar. The rates in this post were not verified against Kaushik's original text.
• Google Search Central, changes to FAQ and HowTo rich results (August 2023).