AI Brand Visibility: Neil Patel Updates Framework – New Tracking Errors to Avoid

Avant de plonger dans les erreurs de tracking, ce chiffre de Neil Patel résume le problème : l’immense majorité des mentions IA sont aveugles.

The error I see on 15 audits every week

I review 15 sites per week. They all have the same problem.

One Tuesday morning, an e-commerce client calls me. 12,400 organic sessions. A catalog of 2,300 product references. And a well-funded AI tracking budget.

« Stéphane, we’re visible in ChatGPT on 14 key queries. Perplexity cites us in 22 conversations. AI Overviews mentions us in 8 snippets. But zero attributed sales. Nothing. »

The silence that followed spoke volumes.

This client didn’t lack visibility. They lacked measurement. Example: Neil Patel, in his 2025 framework update, estimates that 83% of brand mentions in AI responses have no clickable link. The brand is cited, read, maybe remembered… but no clicks land.

And most brand dashboards don’t even see it. We add up mentions. We think we exist. The business doesn’t budge.

This article was born from that realization. It’s not a copy of Neil Patel’s article. It’s the action plan I deployed live so mentions become euros.

What the new Neil Patel framework changes

In 2024, we tracked AI mentions like we tracked Google positions. A prompt became a keyword. A citation, a top 3 ranking. Reassuring, familiar. But missing the mark.

Early 2025, Neil Patel publishes « AI Brand Visibility Tracking: How to Measure What Actually Matters » and gives six keys to rethink measurement. His premise:

« Most teams doing AI brand visibility tracking today have taken a familiar mental model and applied it to an unfamiliar system. Prompts have become the new keywords. Brand mentions, the new rankings. »
— Neil Patel, 2025

In short, we copy-pasted our old SEO reflexes into a conversational environment. The user doesn’t click on a chatbot like on a SERP. They dialogue, bounce around, rephrase. Sometimes they buy two days later, with no direct link.

The six points Neil Patel lists:

• 1. Prompts are not keywords: the same brand surfaces in very different conversations depending on context.
• 2. A mention without a link has little value. You must distinguish citation, clickable link, and transactional page link.
• 3. The tone of the recommendation (neutral or endorsed) strongly changes the impact.
• 4. Each AI surface (ChatGPT, Perplexity, AI Overviews) has its own engagement logic.
• 5. Conversion often stretches across multiple sessions and channels; single-click attribution misses the full picture.
• 6. Measuring only mention volume creates the impression of strong presence when profitability is zero.

These six angles form the basis of the 2025 method. Applying them means stopping the illusions. And starting to see euros, not lines.

Lorsque l’on cesse de mesurer la simple visibilité et que l’on tracke l’influence réelle, les ventes apparaissent. C’est le cas de ce client e-commerce.

First error: confusing « I’m mentioned » with « the customer takes action »

47 sales.

My e-commerce client generated that via AI in January 2025. Seven months earlier: zero.

The breakthrough? We stopped counting mentions. We started tracking influence.

In an AI conversation, the user asks a broad question (« what’s the best [product]? »). The response may mention the brand. If the AI says « some cite X », the user registers it. If it says « X is often recommended for its durability », the user may then search for X on Google, or go directly to the site.

This indirect path, Neil Patel calls the « conversational purchase journey ». It can stretch across two to three sessions, two days. If you have no conversational pixel and no multi-touch attribution, this conversion stays invisible. It falls into the « direct » or « organic » column, never « AI ».

We placed a dedicated tag on the most-cited pages, with a URL parameter (?ref=ai_conversation) assigned to clicks from an identified AI session. Conversions then surfaced: 47 transactions over 31 days, average basket €127, totaling €5,969 in revenue that didn’t exist seven months earlier.

Not a tidal wave. But concrete proof that where the dashboard said « no sales », money was coming in.

Second error: ignoring recommendation strength

Being mentioned is good. Being endorsed is what sells.

Neil Patel distinguishes three levels. Neutral citation (« users mention X »). Moderate recommendation (« X is a good option »). Strong endorsement (« X is widely recognized as the reference »). Only the last shifts purchase intent.

Another client, in the supplements market, experienced this. In November 2024, their brand appeared in 38 Perplexity conversations. Mostly as neutral citations. Clicks? Nearly zero. Sales? None.

We worked on authority signals: placement in deep-dive articles, high-quality third-party citations, Review structured data. In three months, the recommendation flipped. By February 2025, we tracked 12 attributed sales, with a post-impression conversion rate of 3.6%.

The mechanism: stronger recommendation increases indirect click-through rate and intent. You must measure this strength.

I apply the « S » of the DOSE framework here: Structure. We map positive entities associated with the brand in AI conversations (quality, price, customer service). We then structure the site content to reinforce these entities, so as to influence recommendation tone. We track the ratio of endorsed mentions to neutral mentions via conversational listening tools.

An example of results harvested from three clients over January–February 2025:

You read that right: an endorsement generates a conversion rate twenty-nine times higher than a simple citation. If your dashboard doesn’t make this distinction, you risk investing in the left column.

Third error: treating ChatGPT, Perplexity, and AI Overviews as one channel

ChatGPT is not Perplexity. Perplexity is not AI Overviews. Yet many organizations dump their « AI mentions » into the same spreadsheet.

Big mistake.

Neil Patel is clear: the three surfaces drive radically distinct user journeys.

ChatGPT: long conversation, diluted recommendations, few clickable links. Perplexity: concise response, commercial links present, bias toward academic sources. AI Overviews: snippet displayed straight in Google, rarely a link to your brand, but strong influence on the next search.

I saw a client waste 85% of tracked budget by treating all three the same way. They tracked only Perplexity because links were easier to trace. Meanwhile, ChatGPT was generating pre-purchase conversations, invisible in the dashboard.

The fix: distinct tracking per surface. Platform-specific UTMs (?utm_source=chatgpt, ?utm_source=perplexity, ?utm_source=aioverview) and post-click pixel. With that same client, after seven months of segmentation, conversions broke down this way: 62% from ChatGPT, 28% from Perplexity, 10% from AI Overviews.

7 months.

Time needed to understand each surface’s mechanics, configure sensors, and let history accumulate. Today, the marketing team allocates content budget based on these figures, not on whichever surface is easiest to track.

Apply the DOSE framework to your AI visibility: 3 immediate actions

I’m not selling you the method. I’m showing you the pages.

My clients transformed results in seven months flat, with three actions that blend Neil Patel’s method and Guillaume Attias’s DOSE framework (BMO Academy):

• Detect (D). Find the 10 AI conversations where your brand is cited (via Wope, Semji, or your users’ ChatGPT history). Sort by platform, spot associated entities and tone of each mention.
• Structure (S). Create one page per major conversational intent, optimized for brand entity + associated entity. Add Organization, Product, and Review structured data. Goal: give AI models the authority signals to shift from neutral citation to strong recommendation.
• Evaluate (E). Install a conversion pixel on target pages accessible from an AI conversation. Set up a conversational UTM. Track month-to-month the ratio of strong to weak recommendations using multi-touch attribution (Matomo, Piwik PRO). Don’t measure mentions anymore—measure revenue.

Result with the e-commerce client I mentioned earlier: 47 sales attributable to AI in January, €5,969 in additional revenue. No ads.

We stopped focusing on mention curves. We plugged tracking into the till.

What if your current tracking only measures noise?

In reality, many teams think they’re measuring AI visibility. They pile up mentions without distinguishing neutral from endorsed. Without linking it to sales. Without segmenting surfaces.

They’re only measuring noise.

Neil Patel listed six blind spots. I see these gaps every week with my clients. When we fix them, the euros come.

So before requesting another report, ask yourself one question: how many euros did your AI visibility put in the till this month?

If you can’t answer with a number, it’s not your presence that’s the problem. It’s your measurement.

That’s what we fix in one live audit call.

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn