A few months back, someone on a marketing team typed a question into ChatGPT instead of Google, and nothing dramatic happened in that moment; no alert went off and no one noticed anything unusual, but somewhere in that small shift, an entire industry’s assumptions about how people find brands stopped being entirely true.
That’s the strange part about this change, in that there’s no single headline that captures it, because it’s happening across thousands of small decisions made by people who used to type “best running shoes for flat feet” into Google and now ask an AI assistant the same thing and simply go with whatever it tells them.
The New AI Reality in Marketing
The traditional funnel was built on an assumption: awareness leads to interest, interest leads to consideration, consideration leads to a decision, and the brand has a reasonable sense of when each stage happens. It was a tidy way of thinking about the customer journey, and while it was never quite as accurate as the diagrams made it look, it stayed close enough to reality to be useful for a long time.
What’s changed is that the research stage has scattered across several tools at once, and most of them don’t behave like search engines used to. Someone might ask an AI assistant for a quick comparison, scroll through a few reviews, see a forum thread that settles the question for them, and only visit a brand’s website once they’ve already half decided what they want.
This isn’t a guess, since McKinsey’s analysis of personalisation and customer behaviour found that companies that grow faster drive 40% more of their revenue from personalisation than their slower-growing counterparts, which tells you something important: the brands pulling ahead are the ones paying close attention to how people actually move through their decisions, rather than how a textbook funnel says they should. Source: McKinsey, The Value of Getting Personalization Right or Wrong is Multiplying.
For a brand, this changes the job, because it’s no longer enough to rank on a search results page when the content also has to be clear and well-sourced enough that an AI tool can confidently pull from it and represent it accurately when someone asks a related question.
When helping brands restructure their content strategies at Ellipsis Digital, we’ve consistently found that pages built around a specific customer question, written plainly and with a clear answer near the top, tend to perform better across both traditional search and AI-assisted discovery than pages written to sound broadly impressive.
The Funnel Rebuilt for AI
Once the research stage stops following a predictable order, the question becomes what a funnel built around that reality actually looks like, and that starts with how the marketing team itself operates rather than which tools it buys.
From Doers to Directors of Intelligence
There’s a shift happening inside marketing teams that’s easy to miss if you’re not looking closely, where people who used to spend their days writing first drafts, pulling together basic reports or scheduling posts are increasingly spending that time reviewing what an AI tool produced and fixing the parts that are technically fine but say nothing useful.
This isn’t really about job losses, although that’s a separate conversation people are having, and it’s more that the actual skill being asked of marketers has moved. Knowing how to write a good caption still matters, but knowing how to brief an AI tool properly and correct what it gets wrong has become just as important, especially since a lot of AI-generated marketing content right now reads as grammatically correct and completely forgettable, the kind of thing that fills a content calendar without doing anything for the brand.
The teams getting ahead treat that output as a draft that needs a human pass, and that pass is where the brand’s voice survives and where something that sounds like everyone else gets turned into something that sounds like the brand.
Build Systems, Not Isolated Tasks
Here’s something that becomes obvious fairly quickly once a team starts using AI seriously: it works best when there’s structure around it, since a vague prompt produces a vague result every single time.
A team with a shared prompt library, a standard brief format, and a clear sense of what gets automated and what doesn’t will consistently get better output than a team where everyone is improvising from scratch, because the brand voice, the fact-checking step, and the formatting all become repeatable rather than depending on whoever happened to write the prompt that day.
This also speeds up the review step, since if everyone starts from the same brief and the same core prompts, whoever is checking the output knows roughly what to expect and can spot what’s gone wrong much faster, and new team members can learn the system instead of absorbing years of unwritten habits.
Run an AI Literacy Sprint
If all this sounds like a big organisational overhaul, here’s the more manageable version, since one approach that comes up often in how teams are adapting is a short, contained sprint, usually about two weeks, focused on just one or two repetitive tasks.
The process is simple enough in practice. Pick something the team does often and finds tedious, maybe drafting first-pass social captions, summarising long reports, or putting together a weekly performance update, then build an AI workflow for that one task, refine the prompts until the output is genuinely useful rather than just technically complete, and finally share what worked and what didn’t with the rest of the team.
What this does, beyond the time saved on that one task, is build trust in the process. A lot of hesitation around AI tools comes from people imagining worst-case outcomes because they haven’t actually tried it on something low-stakes. A small, well-run sprint replaces that imagined worst case with a real example of what good output looks like, and that spreads through a team faster than any policy memo ever could.
Five Core Capabilities Modern Marketers Need
Across the available research on this shift, five capabilities keep coming up as the ones separating marketers who use AI well from those who just use it.
The first is prompting: giving an AI tool enough context and constraint that it produces something useful on the first or second try rather than the tenth. The second is interpreting data, because AI tools can generate insights quickly, but someone still needs to judge whether those insights actually hold up or whether the model has confidently misread the numbers.
The third is designing simple automations, connecting tools so information flows between them without someone manually copying and pasting. The fourth is blending creativity with AI’s speed, using that speed without letting the output sound interchangeable with everyone else’s. And the fifth, which doesn’t get talked about enough, is judgement: knowing when AI-generated content needs a closer look before it goes out, and when something just feels off even if it’s technically accurate.
None of these are deeply technical skills in the way “learn to code” might have been a decade ago. They’re closer to judgement calls, which is exactly why they’re harder to teach through a course and easier to build through practice. Once a team has these capabilities in place, the next step is applying them at each stage of the funnel, starting with the very top.
AI Visibility Is the New SEO Layer
Google rankings still matter, and for most brands they’ll keep mattering for a long time, but they’re no longer the only scoreboard. AI tools now generate their own answers by pulling from sources they consider clear and trustworthy, and being cited inside one of those answers is becoming its own form of visibility, sitting alongside a traditional search ranking rather than replacing it.
This is often called Generative Engine Optimisation, or GEO, and the practical upshot is that content now has two audiences to satisfy at once: the person reading it, and the AI system deciding whether to reference it. That doesn’t mean writing differently for each one.
It means writing with enough structure, specificity and sourcing that both a human reader and an AI system can quickly understand what the content is about and why it can be trusted. In our work at Ellipsis Digital, we’ve started treating this as a standard part of every content brief rather than a separate workstream, since the same structural habits that help a page rank tend to be the ones that help it get cited.
AI at the Top of the Funnel: Attract
The top of the funnel used to be mostly about keywords and rankings. It still partly is, but there’s now a second audience for that content: the AI systems that summarise and recommend things to people. If a brand’s content is vague, poorly organised, or makes claims without backing them up, it’s far less likely to be the source an AI tool draws from when answering a related question.
What seems to help is content that reads almost like a well-organised reference: clear headers that match the actual questions people ask, definitions stated plainly near the top of a section rather than buried halfway through, and sources cited for anything that sounds like a claim rather than an opinion. It’s not that different from good writing practice generally, but the stakes for getting it right have gone up because there’s now an algorithmic reader as well as a human one.
AI also makes the next step, turning that long-form content into everything else, dramatically faster. A single well-researched article can become a handful of social posts, a short video script, and an email, all without the creative team starting from scratch each time. That matters because audiences are spread across formats, and a brand that only shows up in one of them is leaving a fair amount of visibility on the table.
AI in the Middle of the Funnel: Nurture and Convert
The middle of the funnel is where things get genuinely unpredictable. People don’t move through consideration in a fixed order, and what convinces one person, say a detailed case study, won’t move another person at all, who might respond better to a straightforward comparison page or a short demo video.
This is where AI’s ability to read signals as they happen becomes useful. McKinsey’s research into personalisation found that doing it well can lift revenue by 5 to 15% and improve marketing spend efficiency by 10 to 30% (McKinsey, The Value of Getting Personalisation Right or Wrong is Multiplying). The idea is that instead of sending everyone through the same nurture sequence regardless of what they’ve actually shown interest in, the content adapts based on what someone has clicked, watched or asked about.
There’s a caution worth sitting with here, because it matters. A Gartner survey of 1,464 B2B buyers and consumers found that personalisation generates negative experiences for 53% of customers, who were 3.2 times more likely to regret a purchase and 44% less likely to buy again (Gartner, Personalisation Can Triple the Likelihood of Customer Regret at Key Journey Points). The lesson here isn’t to avoid personalisation altogether. It’s that personalisation needs to feel useful rather than like being watched, and that usually comes down to using the data to help, not just to target more precisely.
AI at the Bottom of the Funnel: Retain and Expand
The bottom of the funnel has always been where the real value sits, in retention, repeat purchases and expansion into other products or services, and it’s also where AI’s predictive ability has one of its clearest, least debatable uses.
Instead of finding out a customer is unhappy when they cancel, predictive models can flag the early signs: a drop in usage, a support ticket that sat unanswered too long, a pattern that historically comes before someone leaves. That gives a team time to actually do something about it, rather than recording it as a loss afterwards.
The same applies to conversion friction, since if a meaningful number of people are dropping off at a specific step in a sign-up or checkout flow, AI tools can help spot that pattern faster than someone manually digging through analytics every week. The value here isn’t that AI fixes the problem on its own, but rather that it points at the problem early enough for a person to actually fix it.
Build an AI-Ready Growth Engine
The last piece of this is less about any single tactic and more about how all of it fits together over time. A funnel built around AI tools generates a steady stream of data: what content got referenced in AI answers, which personalised pages converted better, which retention flags turned out to be accurate.
The brands getting real value from this aren’t necessarily the ones with the most expensive tools. They’re the ones with a habit of looking at that data, writing down what worked, and feeding it back into how the next round of briefs, prompts and workflows gets built.
Over enough cycles, that turns into a system that gets steadily better, not because the AI got smarter on its own, but because the team got better at directing it. It’s the same pattern we keep coming back to with clients at Ellipsis Digital: the brands that review their funnel data monthly, even informally, tend to spot small shifts in behaviour months before the brands that only look at it during quarterly planning.
That’s really the whole shift captured in one thought: the tools changed quickly, and the advantage goes to whoever adapts how they work just as quickly.
Most brands don’t have a traffic problem anymore. They have a visibility problem. Prospects are researching through AI assistants, review platforms and social communities long before they ever visit a website, and if a funnel was built for a different customer journey, it may be time to rebuild it around the one people are actually on.
Key Takeaways
- The funnel is no longer a straight line, so brands need visibility across AI tools, search engines and social platforms, not just the points a traditional funnel diagram used to flag.
- Marketing teams are shifting from doing tasks manually to directing AI systems, which means judgement, review and strategy now matter more than raw execution.
- Content needs clear structure, including direct definitions, sensible headers and credible sourcing, since these are the signals AI tools rely on when deciding what to reference.
- AI visibility, sometimes called Generative Engine Optimisation or GEO, is becoming a layer on top of traditional SEO rather than a replacement for it, and brands need content built for both a human reader and an AI system at once.
- A short AI literacy sprint, where a team rebuilds one or two repetitive tasks around an AI workflow, is one of the fastest ways to build real comfort with these tools.
- The middle and bottom of the funnel benefit most from AI’s ability to read intent signals in real time, letting nurturing and retention respond to actual behaviour instead of a fixed schedule.
- Personalisation helps, but only with restraint, since the evidence shows it can backfire when it feels intrusive rather than useful.
FAQs
Is the traditional marketing funnel completely gone?
Not entirely, but it’s far less linear than it used to be. The stages still exist conceptually: awareness, consideration, decision, retention, but people move between them in any order, across multiple tools, often without a brand noticing until the data shows it.
Do small businesses need to worry about AI visibility?
Yes, arguably more than larger brands, because AI tools tend to favour clear, well-sourced, specific content over generic brand messaging, which is actually an advantage for smaller businesses that can speak with real authority about their specific niche.
What’s a reasonable way to start using AI in marketing without it becoming overwhelming?
Start small and specific by picking one repetitive task, running a short trial with an AI workflow, and judging the results honestly before expanding, since trying to change everything at once tends to produce the kind of flat, forgettable content AI gets criticised for.
Does personalisation always improve conversion?
No, and it’s worth being upfront about this, because personalisation that feels genuinely useful tends to help, while personalisation that feels intrusive or oddly specific can backfire, sometimes quite badly, which is why the goal should be relevance rather than customisation for its own sake.
Will AI replace the need for a marketing team?
The evidence so far points the other way. What’s changing is the kind of work the team does: less manual execution, more review, judgement and strategic direction. The teams getting the most out of AI are the ones treating it as something to manage carefully, not as a replacement for the people managing it.
How do I know if my content is actually being picked up by AI tools?
This is still an evolving area without a standard measurement yet, but checking whether AI assistants reference or summarise your content when asked relevant questions is a reasonable starting point. It’s less precise than traditional search rank tracking, but it gives a sense of direction.
Conclusion
That rebuilding work, thinking about visibility across AI tools, search and social all at once, is where Ellipsis Digital spends most of its time. As a creative and digital marketing agency in Pune, we help brands rework their content and marketing systems around how people actually search, compare and decide today, regardless of whether those people are sitting down the road or halfway across the country. If a funnel feels like it’s losing people somewhere and it isn’t obvious where, that’s usually the first conversation worth having, and we’d be glad to have it with you.
