AI search is rewriting where your traffic comes from. Here's the playbook.

Six months ago, a client we've worked with for years called to ask a strange question. Their organic traffic had dipped 18% over a single quarter. Rankings hadn't moved. CTR hadn't moved. Nothing on the site had changed. Where was the traffic going?

We pulled the obvious reports — Search Console, GA4, paid spend. Everything looked fine. Then we looked at where the answers their prospects were getting were coming from. About forty percent of the questions their best customers had asked us, in sales conversations over the prior six months, were now being answered by ChatGPT — citing competitors.

That's the shift. AI-generated answers are eating the discovery layer of B2B search, and the rules of who gets cited are not the same as the rules of who ranks. This piece is the four-week plan we've been running with clients ever since — including the measurement framework that lets you actually defend the budget.

Why this is happening now, not next year.

Three things changed in the last twelve months. First, the major LLMs went from "answer with confidence, sometimes wrong" to "answer with citations, almost always right." Second, Perplexity, ChatGPT Search, and Google's AI Overviews started converting at rates that justify pricing them as a discovery channel. Third — and this is the part nobody talks about — the same prompts that used to send traffic to your blog post now send a one-sentence summary with a single citation. The click is gone. The brand mention is the thing.

For B2B SaaS in particular, this is closer to a windfall than a problem. The buyers you actually want spend their day asking questions like "what's the difference between X and Y," "how do teams handle Z at scale," and "who's good at this in 2026." Those queries map cleanly onto LLM workflows.

How LLMs actually decide who to cite.

The honest answer is that nobody outside the labs knows exactly. But after auditing roughly forty B2B sites that either are or aren't being cited, four signals come up again and again:

1. 01
   Crawlable, structured answers

   FAQ schema, comparison tables, glossary entries — content where the question is in the markup, not implied by the headline.
2. 02
   Authority signals the LLM trusts

   Mentions on independent sites, podcast transcripts, YouTube descriptions, and forum posts. Backlinks still matter, but the surface is wider.
3. 03
   Recency that's actually visible

   A `dateModified` that the crawler can see, content that updates, version numbers in the URL when relevant.
4. 04
   Clear, opinionated voice

   LLMs prefer text that takes a position. Hedged "some say" copy gets summarized and dropped; specific claims get quoted.

The brands being cited in 2026 will be the ones who, in 2025, wrote like they were teaching, not selling.

The four-week playbook.

This is the sequence we now run on every engagement that has an AEO objective. It's compressed; we've shipped it as fast as nine working days when the source content was already there. None of the steps are individually clever — the compounding is the trick.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "How is AEO different from SEO?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "SEO optimizes for ranking. AEO optimizes for being cited
       by an LLM as the source of an answer. Different signals, different
       measurement, often the same content with different markup."
    }
  }]
}

Measuring something that doesn't show up in GA4.

The measurement gap is the real reason this is hard to sell internally. There is no clean dashboard. What we ship instead is a weekly prompt panel — twenty target queries, run across three LLMs, scored on whether your domain is cited and at what position. Two engineers can stand it up in a day.

Pair that with a simple "How did you hear about us?" field on your demo form. Within six weeks of starting, you'll see "ChatGPT" and "Perplexity" climb past most paid channels for the segments where this works.

The three objections we always get.

• "Won’t this just be a hallucination risk?"
  It can be — if your structured answers are wrong. The whole point of investing here is to be the trusted source the LLM defaults to. Quality of the source content matters more than ever.
• "How does this not just become SEO again with extra steps?"
  Because the unit of measurement is citation, not click. You’ll write fewer, denser, more specific pages — not more thin pages.
• "Our buyers don’t use ChatGPT."
  We hear this and then run a survey. They do. They don’t always tell you they do. Run the survey.

What we'd do tomorrow morning.

If you're reading this and you don't have an AEO strategy yet, the highest-leverage thing you can do this week is pull your last two quarters of sales call transcripts, cluster the questions, and ship one well-structured FAQ page that answers them clearly. That's the whole flywheel start. The rest of the playbook above is just turning the volume up on it.

The window where you can compound here while your competitors are still arguing about whether it's real is roughly twelve months wide. Use it.