For years the web revolved around the simple ritual of typing a query, receiving a page of blue links, and picking one. That habit is still enormous—Google alone processed about five trillion searches in 2024—but it is no longer the only game in town. Conversational AI tools such as ChatGPT, Gemini and Perplexity have added a second pathway: instead of sending people off to hunt for answers, they supply a ready-made response inside the interface. In 2025 the ten largest chatbots logged roughly 55 billion visits, up more than 80 per cent year on year. Search volume stayed almost flat. The message is clear: we’re not watching a replacement; we’re watching a branching.
Why traffic patterns matter to the funnel
Classic marketing funnels assume that awareness, discovery and evaluation happen inside assets you own—landing pages, blog posts, downloadable guides. These stages attract search clicks and pass qualified visitors deeper until the checkout button. Generative answers up-end that order. Awareness and comparison now take place in someone else’s sandbox, often with zero traffic returning to you. When Google’s AI Overviews appear, studies show organic click-through can fall by a quarter or more for that query type. The practical consequence is an inverted funnel: the early conversation moves off-site, while the website becomes a final pit stop for payment or sign-up.
Top of funnel: influence beats clicks
Search engines rewarded meticulous keyword tuning. Answer engines reward authority. Large models lean heavily on sources they view as reliable, including Wikipedia, respected news outlets and sites that publish schema-rich, well-structured content. Your article can be quoted in the summary even though no-one clicks through. That may feel thankless, but it still shapes perception. New success metrics therefore shift from sessions to share of voice inside the answer box—how often the model cites or mentions your brand.
Action items for 2026:
- Write for intent, not for single terms. Cover a topic fully so the model can lift clean sentences rather than fragments.
- Add schema everywhere. Product, FAQ, How-To and Author markup give machines unambiguous signals.
- Produce work that resists compression. Original datasets, interactive calculators and strong first-person stories give users a reason to visit the source.
Middle of funnel: the AI advisor
When somebody asks, “Which heat pump suits a three-bedroom home in Brisbane?” a chatbot compiles specs, reviews and pricing within seconds. The panel it returns feels objective, even if portions are mistaken or out-of-date. That shifts trust from brand sites to the AI itself. If your model fails to list a product, you are invisible at the moment of consideration.
To counter this, marketers need a new craft—call it Generative Engine Optimisation. It combines traditional expertise building with hands-on machine governance:
- Audit the models regularly. Run the key buying questions and record how each platform describes you and your competitors.
- Correct errors at the source. If a model repeats an outdated warranty policy, update every public document that mentions it and lodge feedback with the vendor.
- Strengthen third-party signals. Rich reviews, expert endorsements and government certifications lift confidence scores inside LLMs far more than polished copy alone.
Bottom of funnel: prepare for agents, not just people
Traffic that does reach your site via an AI link converts at a healthy clip. Analytics across several industries show bounce rates drop and time-on-page rises because visitors arrive pre-sold. Yet the bigger shift is still ahead: autonomous agents. By 2026 many consumers will give their assistant standing orders such as, “Re-order my running shoes when the tread depth gets low and the price dips below $150.” The purchase decision, payment and even delivery preference sit in code.
To win in a machine-to-machine market you need:
- Fast, well-documented APIs. If the agent cannot confirm stock or shipping fees instantly, it will pick a rival retailer that can.
- Transparent, structured product feeds. Price, availability, dimensions and sustainability claims should live in a feed that updates in real time.
- Security and privacy assurances in plain language. Developers—and regulators—will punish weak data practices long before a headline scandal hits.
Paid media in an uncertain SERP
Google’s AI panels push some text ads well below the fold. Shopping units often survive near the top, but ordinary search campaigns now face volatile click costs. In response, shift budget to high-intent phrases (“buy solar inverter Brisbane”) and take full advantage of first-party data. Platforms like Performance Max thrive on clean CRM audiences; feed them recent purchasers, and let the algorithm hunt for look-alikes.
Meanwhile, monetisation inside chat interfaces is still experimental. Sponsored recommendations and affiliate links are popping up, yet performance data remain sparse. Keep watch, test early, but avoid locking budget until formats settle.
Brand safety re-defined
Past brand-safety worries focused on ad adjacency. Today the larger risk is the answer itself. Generative systems occasionally hallucinate, and the result can be a phantom product flaw or, as one Canadian airline learned, a discount that never existed. Legal responsibility still lands on the brand.
Build a simple governance loop:
- Monitor: use social-listening style tools designed for chatbots.
- Verify: cross-check flagged statements against official documentation.
- Rectify: publish corrections on authoritative domains and push them to model providers.
Balance this technical vigilance with a human face. Surveys show Australians welcome AI that augments staff but bristle when companies boast about replacing humans.
What next?
Search isn’t dying. It remains the first port of call for quick facts, local coffee shops and urgent “how to” moments. Yet conversational AI has claimed the deeper research territory and is climbing the funnel. Marketers who cling to click counts will watch their reach slip away even while impressions look fine.
Shift the goalposts. Optimise content so machines quote you accurately. Maintain infrastructure that lets agents transact without friction. Audit the answers the public sees, just as you once audited backlinks. By 2026 the winners will be those who treat authority—human and machine-readable—as their primary asset, and clicks as a welcome side effect.
