Framework 3 Layers to Measure Your AI Presence

Why Rankings Are No Longer Enough

A client calls me on a Thursday afternoon. B2B SaaS, $420,000 ARR, 14 people. He tells me: « We’re top 3 on 87 strategic queries. Organic traffic is up. Leads are down. »

I pull up their GA4. +34% sessions over 6 months. −19% qualified conversions over the same period.

The diagnosis takes 20 minutes.

Their Google positions are intact. But 68% of their target queries now generate AI Overviews. ChatGPT mentions 3 competitors in 9 out of 10 prompts. Perplexity mentions them 2 times out of 47 queries tested.

Their « top 3 » measures a reality that no longer pays.

Aleyda Solis publishes in April 2025 a framework in 3 layers to measure what rankings don’t capture: presence in AI Search, the structure that enables it, the business impact that follows. Not a single metric. A stratified system.

Because a user can discover a brand in ChatGPT, form a preference, then type the URL directly into the browser. Zero clicks attributable to AI. 100% influence on the decision.

The classical model measures the click. The new model measures influence, even without a click.

I’m going to show you how to structure this measurement without inventing imaginary attribution.

Layer 1: Presence — Are You Visible, and How?

The first layer measures whether your brand appears in AI responses that matter.

Not « all possible responses. » The responses that match your buyers’ real constraints.

Aleyda Solis recommends pragmatic sampling, not exhaustive coverage. You test 30 to 50 representative prompts, not 10,000 variations.

The 5 presence KPIs she identifiés:

• Mention rate: in how many responses does your brand appear?
• Average position: if it appears, at what rank?
• Citation rate with link: how many mentions include a clickable link to your site?
• Explicit recommendation rate: how often does the AI say « I recommend X » or « X is suited for Y »?
• Representation quality: does the AI describe your offer correctly, or does it invent characteristics?

I deployed this system with 9 clients between January and March 2025. All discovered their presence varied by platform.

Concrete example: tech PR agency, 22 people, UK. Typical prompt: « What PR tool for a B2B SaaS agency of 8 people, $400/month budget? »

Their classical SEO strategy positioned them well on Google. Their presence in ChatGPT was nearly zero.

Perplexity cited their site 7 out of 8 times because they’d published 14 structured case studies with Schema.org and backlinks from TechCrunch, VentureBeat.

ChatGPT preferred brands with more Reddit mentions, Product Hunt, public Slack forums.

Same query. Two different corpora. Two necessary stratégies.

Aleyda Solis insists: build a presence dashboard by platform, then connect it to business questions.

• « Why do our competitors appear 6 times more often in Perplexity? »
• « Why does ChatGPT systematically mention 3 brands, none of which is ours? »
• « When the AI mentions our brand, does it describe the offer correctly? »

Layer 1 doesn’t answer these questions. It makes them measurable.

Layer 2: Readiness — Does Your Structure Enable Visibility?

The second layer diagnoses why your presence looks the way it does.

Aleyda Solis lists 10 structural characteristics of brands visible in AI Search. I give you 7 I observe consistently in my well-positioned clients:

1. Structured content with Schema.org — Product, FAQPage, HowTo, Review.
2. Measurable topic authority — backlinks from sources LLMs index (tech press, academic studies, sector reports).
3. Coherent multi-platform presence — Reddit, Product Hunt, GitHub, Quora, industry forums.
4. Machine-readable product data — prices, specs, comparisons, structured files (CSV, JSON-LD).
5. Long-form content with citations — articles of 2,000+ words citing verifiable external sources.
6. Real FAQs — not keyword stuffing, genuine customer questions with precise answers.
7. Editorial freshness — regular updates, not a blog abandoned for 18 months.

A readiness audit starts from presence results. You don’t do a blind technical audit. You seek why ChatGPT ignores you while Perplexity cites you.

Real case: outdoor e-commerce, 840 SKUs, France. Perplexity presence: 73% on 30 prompts. ChatGPT presence: 11%.

The readiness audit revealed:

• 97% of their product sheets used Schema Product with price, availability, review.
• They had 340 backlinks from outdoor blogs, trail magazines, gear comparators.
• Zero Reddit presence. Zero presence in specialized forums (Hardloop, I-Trekkings).
• No long-form guides citing external sources.

Perplexity scraped their structured data + their press backlinks.

ChatGPT relied on Reddit discussions, forum threads, user comparisons — a corpus they weren’t feeding.

The diagnosis: their technical readiness was strong. Their conversational readiness was zero.

We invested 6 weeks producing 12 guides (« How to choose trail shoes for muddy terrain, » « 5 40L backpacks tested on the GR20 »), structured with HowTo + external citations, then fed 8 Reddit threads with real expérience reports (no spam).

3 months later: ChatGPT presence at 34% on the same 30 prompts.

Aleyda Solis recommends prioritizing actions by effort, not just impact. A TechCrunch backlink has enormous impact, but requires 40 hours of PR outreach. A well-written Reddit thread takes 90 minutes.

I forge readiness roadmaps with an effort/impact table. Always 3 « quick wins » actions (< 10 hours) before heavy lifting.

Layer 3: Business Impact — Does Visibility Create Measurable Value?

The third layer connects AI presence to business results.

Aleyda Solis identifiés 4 confidence levels for measuring business impact:

1. Observed impact — directly attributable data (clicks from Perplexity, traffic from ChatGPT via referrer).
2. Proxy signals — indirect behaviors (branded search increase, direct traffic up, branded CPC down).
3. Modeled impact — probabilistic attribution (post-purchase surveys, multi-touch models, holdout tests).
4. Declared impact — what customers say (« I discovered your brand via ChatGPT »).

She stresses: don’t mix the 4 levels into a single number. A Perplexity click is observable. ChatGPT influence without a click is modeled.

I build impact dashboards with 3 distinct columns:

Real case: SME accounting SaaS, 680 paying customers, Belgium. We deployed the 3-layer framework in October 2024.

Layer 1 (Presence): ChatGPT mention rate 41%, Perplexity 67%, Gemini 29%.

Layer 2 (Readiness): audit revealed zero comparative content, generic FAQs, incomplete Schema.

Actions: 9 comparative guides (« What accounting for Belgian freelancers, » « Accounting SaaS vs Excel »), FAQs restructured, Schema FAQPage + HowTo deployed on 47 pages.

Layer 3 (Impact), 5 months after:

• Observed clicks from Perplexity: 890/month (vs 120 before)
• Direct traffic: +52%
• Branded Google searches: +41%
• Branded CPC: −23% (less competition on their name = brand awareness signal)
• Post-signup survey: 22% declare « discovered via ChatGPT or Perplexity »
• MMM model (Marketing Mix Modeling): AI Search contribution estimated at 14-21% of new signups

Revenue attributable with high confidence: $8,400/month (observed clicks × conversion rate × average basket).

Estimated total revenue (model + proxy): $18,000-26,000/month.

I never tell them « AI Search generates $26,000. » I tell them: « $8,400 is observed. $18,000-26,000 are modeled with a ±30% error margin. »

Aleyda Solis hammers home this point: don’t overstate attribution. AI Search often influences without leaving a direct trace. Measure what you can observe, estimate the rest carefully, never mix the two.

Connecting the 3 Layers: From Report to Decision System

The framework only works if the 3 layers talk to each other.

Low presence → you go into Readiness seeking why.

Strong readiness but low presence → you verify if platforms actually index your content (sometimes they don’t).

Strong presence but zero business impact → you diagnose if you appear on the right prompts (maybe you’re visible on informational queries, not transactional ones).

Aleyda Solis gives an example with Finchling (keyword research tool). Their ChatGPT presence was 67% on 40 SEO tool prompts. Observed business impact: 340 clicks/month, 12 trial conversions.

Readiness audit: structured content ✓, press backlinks ✓, Schema ✓, weak Reddit presence.

Diagnosis: they appeared on generic prompts (« SEO tools »), not intent prompts (« keyword research tool for local SEO agency »).

Action: rewrote 6 product pages to include specific use cases (« local agency, » « B2B SaaS, » « e-commerce »), added 14 customer case studies with Schema.

4 months later: presence on intent prompts +290%, trial conversions ×2.4.

The framework transformed « we’re visible but it doesn’t convert » into « we’re visible on the wrong prompts, here’s how to fix it. »

I deploy this system with minimal viable setup:

• Presence: 20-30 representative prompts, tested manually 1×/month on 3 platforms (ChatGPT, Perplexity, Gemini).
• Readiness: 10-criterion checklist (Schema, backlinks, FAQs, long-form, forum presence).
• Impact: 3 observed metrics (GA4 clicks, direct traffic, branded searches) + 1 quarterly survey (« How did you discover us? »).

Time required: 4-6 hours/month to maintain the system once installed.

You don’t need a real-time dashboard with 47 metrics. You need a system that connects visibility → structure → revenue, and diagnoses where it breaks.

Where to Start: 4-Step Setup

If you’re starting from zero, here’s the sequence I deploy with my clients.

Step 1: Map 20 Critical Prompts

Not « all possible prompts. » The 20 prompts your buyers actually type.

Method: take your 10 top SEO landing pages. For each, write 2 conversational prompts a user would ask ChatGPT.

Example outdoor shoe e-commerce:

• Landing page: « Trail shoes for muddy terrain »
• Prompt 1: « What trail shoes for running muddy terrain in autumn? »
• Prompt 2: « Waterproof trail shoes with good grip, budget €150 max »

You get 20 prompts. You test them manually on ChatGPT, Perplexity, Gemini. You note if your brand appears, at what rank, with or without link.

Time: 3-4 hours.

Step 2: Readiness Audit on 10 Criteria

Checklist I use:

1. Schema.org deployed on product/service pages? (Product, FAQPage, HowTo)
2. Real FAQs (not keyword stuffing)?
3. Long-form content with external citations?
4. Backlinks from tech press / studies / reports?
5. Reddit / industry forum presence?
6. Comparative guides (« X vs Y, » « How to choose Z »)?
7. Structured product data (price, specs, stock)?
8. Editorial freshness (content updated < 6 months)?
9. Pages with 2,000+ words including lists, tables?
10. Brand mentions on conversational platforms (Reddit, Quora, forums)?

Score: 1 point per criterion validated. < 4/10 = weak readiness. 7-10/10 = strong readiness.

Time: 2-3 hours.

Step 3: Measure Observed Impact

Configure GA4 to track:

• Clicks from Perplexity (referrer « perplexity.ai »)
• Direct traffic (month-by-month evolution)
• Branded Google searches Search Console

Add a question to your lead/signup form: « How did you discover us? » with « ChatGPT / Perplexity / other AI » option.

Time: 1 hour initial setup.

Step 4: Connect the 3 Layers in a Single Doc

Google Sheet, Notion, whatever. Structure:

You update this table 1×/month. You compare evolutions. You diagnose where it gets stuck.

Time: 30 minutes/month once in place.

No magic tool. No continuous scraping. A lightweight system connecting visibility → structure → revenue.

What This Framework Is Not (and Why It Matters)

Aleyda Solis clarifies 3 limits I systematically outline with my clients.

1. It’s Not a Complete Attribution System

You’ll never be able to say « AI Search generated exactly $47,340 this quarter. » Part of the influence remains unmeasurable.

A user reads a ChatGPT response, remembers 2 brands, searches Google 3 days later, clicks a paid ad, converts. Which channel drove the decision?

The framework measures what’s observable, estimates the rest, but doesn’t invent certainty where there isn’t any.

2. It’s Not a Substitute for CRM or Product Analytics

The framework diagnoses if your AI presence influences decisions. It doesn’t replace your cohort tracking, churn analysis, existing multi-touch attribution.

It compléments. It doesn’t overwrite.

3. It’s Not a Guarantee That Visibility = Revenue

You can be cited 80% of the time in ChatGPT and generate zero conversion if:

• You appear on informational prompts (« How does X work? ») not transactional (« What tool X for Y? »)
• The AI misdescribes your offer (invents features you don’t have)
• Your site doesn’t convert traffic once it lands

Layer 3 (Business Impact) diagnoses whether presence converts to value. If it doesn’t, you climb back through layers 1 and 2 to understand why.

It’s a diagnostic system, not a magic formula.

I see too many dashboards mixing impressions, clicks, estimated conversions, modeled influence into a single « AI Search ROI » KPI. That’s an illusion of precision.

Aleyda Solis’s framework does the opposite: it separates what’s measurable from what’s estimable, and diagnoses where it breaks.

How I Deploy This Framework With Clients

I don’t sell an AI Search audit. I deploy this framework in 3 phases over 8-12 weeks.

Phase 1: Presence Diagnosis (weeks 1-2)

• Map 20-30 critical prompts
• Manual tests on ChatGPT, Perplexity, Gemini, AI Overviews
• Presence table by platform
• Identify gaps (« Why does Perplexity cite us 9×, ChatGPT 0×? »)

Phase 2: Readiness Audit (weeks 3-5)

• 10-criterion structural checklist
• Competitive analysis (3-5 well-visible brands: what do they have you don’t?)
• Roadmap actions prioritized by effort/impact
• 3 quick wins identified (< 10 hours each)

Phase 3: Impact Setup + Deployment (weeks 6-12)

• GA4 tracking configuration (AI referrers)
• Post-conversion survey
• Deploy 3-5 priority readiness actions
• Measure observed impact at 30, 60, 90 days

Final deliverable: single doc connecting presence → readiness → impact, with 4-6 hour/month maintenance dashboard.

Client investment: $8,000 to $14,000 depending on catalog size / prompt count / readiness complexity.

What differentiates this from a classical SEO audit: we don’t deliver 200 pages of technical recommendations. We install a system that continuously diagnoses where you lose visibility and why.

The first call is a live audit. I test 5 prompts on the spot. You see where you appear. You see where competitors appear. We diagnose the structural flaw in 45 minutes.

No PDF. Pages that move.

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