AI searchAI searchOrganicMeasurement

Your traffic is falling and your pipeline is fine. That is the story

Avishai Sam Bitton 5 min read

The argument

Sessions were always a proxy for being found. The proxy broke, the underlying thing did not, and most teams are panicking about the wrong number.

I have had four versions of the same call this year. Organic sessions are down twenty to forty percent year on year. Rankings are broadly stable. Pipeline is flat or slightly up. Somebody wants to know what broke.

Nothing broke. The top of the funnel moved into a place where sessions are not generated. A buyer who would once have clicked four blue links now asks an assistant, gets a synthesised answer with three brands named in it, and only visits a site when they have already decided who is worth visiting.

Which pages lost the traffic

The decline is not evenly spread, and the shape of it is the diagnosis. Split your organic pages into two groups and look again.

Informational pages, down heavily

  • Definitions, 'what is' explainers, basic how-to content
  • Questions with a single correct answer that a model can state without a click
  • Long tail queries that used to arrive one visit at a time

Commercial pages, broadly stable

  • Pricing, comparison and alternatives pages
  • Product and integration pages tied to a specific job
  • Anything a buyer opens to verify a claim before spending money

Verdict: If your loss is concentrated in the left column and your pipeline is unchanged, you lost traffic that was never going to buy. That is a reporting problem, not a demand problem.

The genuinely worrying pattern is the inverse: commercial pages declining while informational pages hold. That usually means you are being left out of the shortlist an assistant produces, and the buyer is arriving at a competitor's comparison page instead of yours.

Losing traffic that never converted is not a crisis. Losing the shortlist is, and it does not show up in a sessions chart.

The measurement that replaces sessions

  1. 1

    Prompt coverage

    Write the twenty questions a real buyer asks in the ninety days before they choose a vendor in your category. Run them across the major assistants on a fixed schedule and record whether you are named, what you are named for, and which source the model leaned on. This is the closest available equivalent to a rank tracker.

  2. 2

    Branded demand

    Branded search volume, direct traffic and the share of demo requests that arrive already naming you. When answer engines do their job, they produce name recall rather than clicks, and this is where it lands.

  3. 3

    Self reported attribution

    One free text field on the demo form. It is imprecise, unmodellable and by far the most honest data you will collect. When buyers start writing 'ChatGPT suggested you' in that box, you have your answer.

What I would not do

  • Do not publish more definitional content to win the traffic back. That is exactly the content the assistant absorbed, and republishing it feeds the thing that replaced you.
  • Do not block the AI crawlers in retaliation. You cannot be cited by a system that is not allowed to read you, and the traffic does not come back either way.
  • Do not set a target to restore last year's session count. You will hit it with traffic that costs money and produces nothing.

The uncomfortable part of this shift is that it removes a number that made marketing legible to people outside marketing. Sessions were easy to explain and easy to trend. What replaces them is messier: a coverage score across prompts, a branded demand line, and a stack of verbatim quotes from buyers.

Messier and closer to the truth is a good trade. It was always a proxy. The job was never traffic, it was being the name that comes up when someone with a budget asks who solves this.

What I would do Monday

  1. 1Segment organic traffic into informational and commercial pages before you interpret any decline.
  2. 2Add a self reported source field to your demo form and read the verbatims weekly.
  3. 3Run your twenty highest intent buyer questions through three assistants and log who gets named.
  4. 4Stop reporting total sessions as a headline metric and replace it with branded search plus hand raisers.

Who wrote this

Avishai Sam Bitton

Founder, DemandBox

Avishai runs demand generation programs for B2B SaaS companies across performance marketing, SEO, and answer engine optimization. He works directly with the teams he advises, with no account managers in between.

Connect on LinkedIn

The long version

What Is Answer Engine Optimization (AEO)? The Complete 2026 Guide

AEO is the work of getting cited inside AI answers. How each engine picks sources, the four step citation pipeline, how AEO differs from SEO, and how to measure it.