Engineering for Answer Engines: Why Your Organic Traffic Was Redistributed, Not Lost

Introduction

Two out of three Google searches in the US now end without a click, and most organic programs were built to deliver traffic, not answers. Jordan Koene, CEO and Co-Founder of Previsible, has spent his career on the enterprise side of this problem, first ru ing eBay's largest traffic cha el as Head of SEO and later as CEO of Searchmetrics. Today his consultancy helps Fortune 100 brands prepare for AI-driven discovery, and he has a blunt read on the 30% organic declines CMOs are reporting. The intent didn't disappear. It moved somewhere your dashboards can't see.

n

The Traffic Went to the Response Layer

Koene's core argument is that we have shifted from a "rankings and choice" era to a response-driven one. Users still want shoes, groceries, and CRM recommendations, but they increasingly get those answers from a chatbot instead of sca ing ten blue links. As he put it, "users' intent haven't changed — it's really the medium by which they consume the information that has changed." The organic cha el hasn't lost relevance. It has lost the one signal marketers knew how to report on.

n

Why This Is a Measurement Crisis

Koene calls this the biggest problem in digital measurement right now, because it hits the one cha el nobody pays for. Paid media buys you a performance output. Organic value now scatters into direct traffic, null attribution, and AI referrals that are far too small to explain the gap. When leadership asks what's working, the old model has no honest answer.

n

Replacing the Click as a Success Metric

Koene's replacement metric is impression value. Google and Bing now report how often AI-generated results use a specific page, which gives you a page-level view of visibility before any click occurs. That matters because Google's AI Overviews and AI Mode reach close to 3 billion monthly users, roughly three times ChatGPT, and run on the same underlying model technology. If you only have one data source, Google's is the benchmark.

n

Treat Visibility Scores as Synthetic

The third-party visibility tools are useful but should be read with caution. ChatGPT, Perplexity, and Anthropic release no impression data, so these tools score the prompts you chose to track, not what users actually typed. Model releases arrive three to four times a year and reshuffle the deck the way Google algorithm updates once did. Koene suggests supplementing scores with pull data: which pages LLM crawlers fetch, and how often. A page that bots return to repeatedly is being used to answer someone's question.

n

Which Models Actually Matter

Previsible's a ual AI traffic study across more than 100 websites found ChatGPT drives over 90% of pure LLM referral traffic, with Gemini gaining fast. But the models are verticalizing. ChatGPT skews consumer, Claude skews B2B and developer research, and Perplexity's financial partnerships pull its traffic toward finance. Partnerships shape which audience sits behind each screen, so your ideal customer profile should dictate where you focus.

n

Optimizing Without Chasing Hype

Koene's first warning is to ignore the hype cycles. Reddit went from everywhere in AI responses to single-digit presence within six months, and brands that had never used the platform wasted budget flocking there. The durable play is understanding your brand's presence across the sources LLMs cite. For B2B, that means current ratings on G2 and Capterra, presence in relevant listicles and trade media, and using citation capture in your tracking tools to find exactly which sources shaped a given answer.

n

On-Page Work Still Counts

For sites that get heavy LLM crawl activity, structured data, FAQ blocks, clean heading hierarchy, and clear summaries act as quick cues models retrieve to build a response. Sometimes that means a stripped-down HTML version for bots, or building API documentation in a separate stack. Koene expects agent-to-agent workflows, where a user's agent completes tasks on your site, to become a standard development layer within a year. One AI startup founder asked him for SEO help while blocking search engines entirely. Accessibility is the first check.

n

Brand Perception Is the Real Signal

Brand marketing has modest direct influence on AI visibility. Brand perception, the accumulated sum of reviews, support responses, employee posts, and community chatter, has a lot. Reputation prompts like "most trusted" or "easiest to use" cause LLMs to leave your site and source from third parties. Meanwhile, About Us and pricing pages across Previsible's 70-plus clients are seeing significant increases in LLM crawl and traffic for the first time in years. Stale pricing confuses models and humans alike, so clean up what you control first.

n

Conclusion

The click is dying, and no single metric replaces it. Measure impressions where Google provides them, treat visibility scores as directional, watch crawl frequency, and audit the third-party sources models actually cite. Refresh About Us and pricing pages, keep review profiles current, and build from cha els that already work rather than chasing the next Reddit. Billboards and podcasts still count, but only when they generate digital conversation that models can read. Organic performance was redistributed, and the marketers who win will be the ones who learn to see where it went.