Marketing Funnels in an AI-Answer World

#01What is the classic marketing funnel?

The common simplified model:

Our view: the funnel is a simplification. It's useful for planning, but real buyers rarely move in a straight line.

#02What is the "messy middle"?

Google's consumer insights team described the space between a trigger and a purchase as the "messy middle", where customers are won and lost (Think with Google). Their research says people switch between two mental modes: exploration, an expansive activity of looking for information about a category's products and brands, and evaluation, a reductive activity of weighing the options. They loop through these modes repeatedly, across search engines, social media, aggregators and review sites (Think with Google).

Our view: AI assistants compress this loop. A buyer can explore ("what tools exist for X?") and evaluate ("compare the top three for a 10-person team") in one chat.

#03How does AI change each stage?

Google says AI Overviews help people get the gist of complex topics and AI Mode suits questions needing further exploration, reasoning or comparison, with links to supporting sites (Google). Our view of the effect by stage:

#04How do you map funnel stages to prompts?

Fictional prompt set for a project management tool:

Run these in AI assistants and note whether you're mentioned at each stage. Our view: many brands do well on awareness content but disappear at consideration, where shortlists are made. See building your first prompt set.

#05What should you measure at each stage?

Google counts AI Overview and AI Mode appearances inside Search Console's "Web" performance data rather than separately (Google), so pair Search Console with direct AI answer checks. See tracking AI traffic in GA4 and measuring AI share of voice.

#06Why ask customers how they found you?

Some AI influence leaves no click. A buyer may read an AI answer, then search your brand name or type your URL. Our view: add "How did you hear about us?" with an AI assistant option to sign-up or demo forms. See self-reported attribution.

#07How should you prioritise across the funnel?

Our view: start where buyers make shortlists. Run a small prompt set covering all four stages, then fix the stage where you're most absent. For most B2B brands that's consideration: category questions such as "what tools exist for…" where competitors are named and you aren't. The fixes there are usually honest comparison pages, clear category positioning on your own site, and mentions on the review sites and communities that AI answers draw on. Awareness content is easier to produce but rarely the gap. Re-run the same prompts monthly so you can see whether each fix moved the stage it targeted, and keep the prompts fixed so results stay comparable.

#08Where do teams go wrong?

• Treating AI as only a top-of-funnel channel. Comparisons and recommendations happen mid-funnel.
• Ignoring third-party sources. Google's research names aggregators and review sites as part of the loop (Think with Google).
• Relying on last-click attribution alone. It misses AI answers that influence without a click.
• Letting product facts go stale. Wrong prices in AI answers hurt at the decision stage.

#09Where next?

Read the buyer journey inside ChatGPT, brand positioning basics and UTM parameters and attribution basics.

#10From MentionX data: which funnel stages mention brandsOriginal data

Method. Read-only aggregate queries over MentionX scans from 22 August to 29 October 2026 (UTC): 2,217 engine runs (2,154 completed) across 18 anonymised projects and six engines.

Prompt intent lines up roughly with funnel stages. In the MentionX sample, brand mention rates differ sharply by intent:

Our view: problem prompts, the top of the funnel, rarely name a brand, while comparison and brand-specific prompts, closer to the decision, almost always do. Content for each stage should match that: teach at the top, compare honestly in the middle and be accurate about yourself at the bottom.

Limits. A convenience sample of mostly B2B software projects, run through engine APIs, which can behave differently from the consumer apps. Perplexity, Copilot and DeepSeek figures come from a single day (26 October 2026), so treat them as a snapshot, not a trend.

#11Worked example: Ledgerline maps its funnel to prompts

Ledgerline is a fictional accounts-payable tool. Figures are illustrative.

Ledgerline's marketing team lists 40 prompts and tags each with a funnel stage. Awareness prompts ask how to reduce late invoice payments. Consideration prompts ask for the best accounts-payable tools for mid-sized firms. Decision prompts compare Ledgerline with competitors A and B, or ask about its pricing.

Monthly runs show Ledgerline is often named in decision prompts but rarely in consideration prompts. The team writes a buyer's guide and gets listed on two independent review sites, then tracks the consideration prompts for a quarter. All figures and prompts are illustrative.

#12Does the funnel still matter at all?

Yes, as a model of what buyers need at each point. What changes is where they get it. A buyer might go from a problem to a shortlist in one AI conversation. Our view: the stages still exist, but they can happen in minutes rather than weeks.

#13What content suits each stage?

Awareness: clear explanations of problems and how to solve them. Consideration: honest guides, comparisons and third-party reviews. Decision: accurate pricing, features, integrations and case studies. Retention: docs and support content that keep customers successful.

#14How does attribution change?

AI answers often don't send a click, so traffic data underestimates their role. Combine referral data, prompt tracking and self-reported attribution, such as a "how did you hear about us?" field. None is perfect alone.

#15How do you report funnel performance?

Show mention rate and citations by stage, alongside pipeline and revenue. Highlight the stage with the biggest gap and the plan for it. Keep the report short.

#16Who owns each stage?

Content teams often own awareness and consideration, product marketing owns decision content, and support owns retention docs. Agree owners so gaps don't fall between teams.

#17How often should you revisit the mapping?

Quarterly. Buyer questions change, and new prompts appear as your market evolves.

#18How do AI answers affect the awareness stage?

At the awareness stage, buyers ask about problems, not products. An AI answer might explain why invoices get paid late and list common fixes without naming any vendor. Your aim is for your explanation to be the one the answer draws on, even when your brand isn't named. Write clear, practical guides on the problems your product solves, with definitions, causes and steps. Our view: being the cited source for a problem builds familiarity before a buyer ever searches your name.

#19How do AI answers affect consideration?

During consideration, buyers ask for shortlists: "best tools for X" or "alternatives to Y". Engines often name several brands and draw on review sites, comparison articles and community threads. This is where third-party sources matter most. Make sure you're listed accurately on the directories and review sites engines cite, and publish honest comparisons that explain who each option suits.

#20How do AI answers affect the decision stage?

At the decision stage, buyers ask specific questions: pricing, integrations, security, contract terms. Engines answer from your own pages if they can find and read them, and from third parties if not. Keep pricing, features and policies clear, current and crawlable. A wrong answer here can lose a deal that was nearly won.

#21What about post-purchase?

Customers ask AI tools how to use products, fix problems and get more value. Clear docs and help articles help them succeed, and they also give engines accurate material about what your product does. Retention content is often overlooked in funnel planning, but it shapes how your product is described.

#22How do you build a funnel prompt set?

Start with real questions from sales calls, support tickets and search data. Tag each by stage and intent. Aim for a balanced set, with enough prompts at each stage to see patterns. Review it quarterly and retire prompts that no longer reflect how buyers ask.

#23How do you explain this to leadership?

Show one example prompt per stage, what the engine answered and whether you appeared. Then show the stage-level mention rates and the plan for the weakest stage. Concrete examples persuade faster than abstract frameworks.

#24Does the funnel differ by engine?

Engines vary in how often they name brands and show citations, as the MentionX table above shows. A stage where one engine rarely names brands might be very different in another. Track the engines your buyers actually use.

#25How do you prioritise when budget is tight?

Pick the one stage with the largest gap and a realistic fix, such as consideration prompts where accurate listings and one strong comparison guide could help. Do that well for a quarter before spreading effort across every stage.

#26What's a good first step this month?

Tag your existing prompts by funnel stage and calculate the mention rate for each. That single table usually shows where to focus, and it takes an afternoon. Share it with sales and ask whether it matches what they hear from buyers.

#27Common questions

Is the funnel dead? No. Buyers still move from problem to decision; AI answers just shorten the path.

Which stage should we start with? Usually consideration, where engines recommend brands and you can influence sources.

How do we measure awareness? Track problem prompts for citations of your content, even when you're not named.

Does this apply to B2C? Yes, though B2C journeys are often shorter and more impulse-driven.