There’s no gentle way to say it. Top-of-funnel SEO is getting eaten alive.

The culprits? AI-powered summaries, zero-click searches, and a new generation of privacy-first browsers that obscure attribution. Ask ChatGPT (or LLM of your choice) how to choose a CRM, and it’ll serve up a helpful guide in seconds. Ask Google the same thing, and you’ll soon get an AI-generated overview before even seeing a blue link. Either way, your lovingly crafted blog post, if it ranks at all, is a sideshow.

The first click now often happens elsewhere. And the rest of us? We’re just winking in the dark.

The death of TOFU, and what comes next

The SEO playbook of the past decade was simple: cast a wide net. Write comprehensive guides, answer beginner queries, and pull in traffic from curious users. That strategy still has its place. But in an AI-first world, the top of the funnel (TOFU) is collapsing. AI eats those informational searches, regurgitates them with polite confidence, and leaves the rest of us fighting over scraps.

This isn’t just a visibility problem. It’s a measurement, one too. Attribution has become a guessing game. Server-side tracking is patchy. GA4 reports often show ‘Direct’ traffic as the biggest channel, but it’s rarely true. On one Shopify report we audited, 70% of ‘new customers’ had no referrer at all. ‘None’ was the source. Not exactly actionable insight.

So we’ve changed tack. We’re guiding clients to look further down the funnel — to the conversion point, and work backwards from there. It’s not as flashy as 10x-ing your content with AI. But it’s grounded in commercial truth.

A new approach: tracing bottom-of-funnel conversions

Here’s the methodology. It’s simple, practical, and scalable:

1. Use GA4 to identify First-Time Purchasers (FTP)

Focus on users who’ve completed a transaction, ideally, those whose conversion paths are short and strong.

2. Isolate the landing page

Use GA4’s “Landing Page + Session source/medium” dimensions to understand where the purchase journey started.

3. Reverse engineer the query

This is where AI comes in. Once you know the landing page, you can prompt ChatGPT (or your tool of choice) to identify bottom-of-funnel (BOFU) search queries that likely drove the visit. These are the keywords with commercial intent. They’re rarely poetic. They’re often clunky. But they convert.

Let’s say your FTP landed on a product page for a CRM tool. Using ChatGPT, you might input:

“What search queries would lead a buyer to land directly on this CRM pricing page?”

The model might return:

• “best CRM for small business with Outlook integration”

• “affordable CRM tool UK”

• “buy CRM with free trial”

Now you’ve got a hit list of high-intent phrases — real bottom-of-funnel queries. These are the keywords that matter. Of course, in the case of a SaaS provider there would be a strong argument for server side tracking to enable them to also analyse this data for the purposes of attribution.

Why this works

There are three reasons this approach is effective:

1. It reclaims insight from broken attribution

GA4 isn’t perfect, but starting with a known conversion and working backwards gives you a cleaner picture than guessing where an ‘awareness stage’ user might go next. The fog clears when you follow the money.

2. It prioritises commercial intent

Not all traffic is equal. Someone searching “what is a CRM?” isn’t ready to buy. Someone searching “CRM with Outlook plugin buy now” probably is. BOFU terms still break through the AI noise because the stakes are different — the searcher isn’t just browsing. They’re buying.

3. It moves SEO from a volume game to a value one

We’ve seen too many teams still measuring performance by sessions or impressions. That’s yesterday’s metric. Today’s challenge is to connect SEO to revenue. This methodology does that. It gives marketers a way to trace clicks that count.

TOFU might still matter — but not like it used to

There’s still a role for TOFU content. Brands with budget and patience can build awareness, capture email signups, and nurture leads. But the cost-to-reward ratio is shifting. Informational content is now a playground for AI and aggregators. You can rank, sure. But you’ll rarely win.

What’s more, TOFU content can cloud judgement. Vanity metrics inflate. Sessions go up, but conversions stay flat. Marketing teams scramble to explain the ROI. Meanwhile, the BOFU queries — the ones that show up when someone’s ready to buy — get ignored because they’re too “low volume.”

Volume doesn’t pay the bills. Value does.

The infrastructure challenge

This shift isn’t just about keywords. It demands infrastructure.

Attribution must be cleaner. That’s why we recommend tools like Elevar for eCommerce clients. It syncs server-side events with marketing platforms and clears up the mess GA4 often leaves behind. Without it, you’re flying blind.

It also means changing how you brief your writers, your SEO team, even your AI assistants. Instead of asking:

“What can we rank for?”

Ask:

“What would someone type when they’re seconds away from buying?”

Then make sure you’re the brand they find.

What this means for strategy

Over the next 12 months, we’ll be applying this methodology across sectors. We’ll share what we learn — what works, what doesn’t, and how AI tools can support (not supplant) this process. We’ll explore how to:

• Tag and track BOFU landing pages

• Train LLMs on your own conversion data

• Build content around intent, not just keywords

• Measure SEO performance by revenue, not reach

And yes, we’ll probably post more pictures of tofu. It’s a visual metaphor now. And our Creative Director eats a lot of it.

Final thought

In the words of a (slightly reimagined) David Ogilvy:

“Optimising for TOFU in an AI-first world is like shouting into Google. You’re making noise, but the buyer’s already had their answer — and bought from your competitor.”

It’s time we all stop shouting.

And start listening to what the bottom of the funnel is telling us.

‍