The old SEO scoreboard is breaking. For Amazon-first consumer brands, the metrics that matter now are prompt presence, citation rate, and AI shortlist share. Here's what to retire, what to track, and how to tie it back to revenue.
Search Engine Land's take is simple: the old SEO scoreboard is breaking. If customers are getting answers inside AI, the metrics that matter are prompts, mentions, and citations — not just traffic and CTR.
For consumer brands selling on Amazon, this matters even more. AI discovery happens upstream, while conversion often happens on Amazon. So you need a measurement stack that connects [AI visibility](/blog/ai-search-101-2025) → buyer intent → Amazon outcomes.
This post lays out what to retire from your dashboard, what to track instead, and a side-by-side that maps the swap.
## Metrics to retire (or demote hard)
**Keyword rankings as the main KPI.** Rankings still help, but they don't tell you if you're being recommended in AI answers. AI can summarise the category without ever rewarding your number-three position. We've covered the [rank-to-sales decoupling](/blog/ai-search-shrinking-ranking-decides-sales-2025) in more depth.
**Top-of-funnel organic traffic volume.** Traffic is easy to celebrate and hard to cash. Visibility and outcomes matter more than raw visits in a zero-click, AI-heavy world. A 20% drop in organic traffic with stable revenue isn't a problem — it's a sign AI is doing the qualification for you.
**Vanity impressions and CTR.** Useful as diagnostics, weak as business metrics. They don't tell you if shoppers are choosing you. CTR especially has lost its meaning in AI-blended SERPs.
**Generic "AI visibility" scores with no receipts.** If a tool can't show which prompts, which answer, and whether you were cited or just mentioned, it's not something you can operate against. The number is impressive on a slide. It's useless in a strategy review.
**"SEO traffic" as the north star.** For Amazon-first brands, the scoreboard is Amazon sessions, conversion, contribution margin, review velocity. SEO traffic is supporting signal, not a goal in itself.
## Modern metrics to track (what actually moves revenue)
The Search Engine Land guidance keeps circling the same pillars: track brand presence in AI answers, citations, and visibility across AI platforms. Here's how to translate that into an Amazon-brand dashboard you can actually run against.
### 1. Prompt monitoring (your new "rank tracker")
Create a monthly prompt set of 25–50 real buyer questions. The questions should look like:
- *"Best [category] for [use case]"*
- *"[ingredient or material] safe for [audience]?"*
- *"[your brand] vs [competitor]"*
- *"What size [product] should I buy?"*
- *"Does [product] work with [compatibility]?"*
Run them across ChatGPT, Gemini, Perplexity, Claude, and [Rufus](/blog/what-is-amazon-rufus-2026). Score each response 0 (absent), 1 (mentioned), 2 (top recommendation). Track the trend monthly.
### 2. Brand mention rate in AI answers
Out of your prompt set, what percentage of answers mention your brand by name? Break it down two ways:
- **Overall mention rate.** Across the full 25–50 prompts.
- **High-intent mention rate.** Filtered to comparison, best-for, and compatibility prompts — the ones closest to a purchase decision.
The high-intent rate is the metric your CFO actually cares about, even if they don't know it yet.
### 3. Citation rate
When AI provides sources or links (which is increasingly the norm in Perplexity and AI Overviews), how often does your site get referenced? Track citation rate and which pages get cited.
This is where the [Schema.org and structured data work](/blog/what-schemaorg-foundational-guide-llm-optimisation-2026) shows up in the dashboard. Pages with cleaner schema get cited more.
### 4. Recommendation inclusion rate (AI shortlist share)
For "best of" prompts, log whether you appear in the shortlist and how often you're top 1–3. Break it down by category segment — "for sensitive skin", "budget", "premium", "travel" — so you can see where you're winning and losing.
### 5. Higher-intent engagement that ties to Amazon outcomes
This is the bridge between AI visibility and money. Don't just measure sessions. Measure behaviour that predicts conversion.
- Clicks to Amazon (use Amazon Attribution where possible).
- Storefront visits.
- Branded search lift on Amazon (proxy for demand creation).
- PDP conversion rate and review velocity trends.
When AI visibility moves up, branded search on Amazon usually moves up 2–4 weeks later. That correlation is the measurement bridge.
## The retire-vs-track table
| Retire | Why | Track instead | Why it matters |
|---|---|---|---|
| Keyword rank position | AI compresses 10 results to 3 | Prompt presence | What the model picks now decides sales |
| Total organic traffic | Zero-click is by design | Mention rate on high-intent prompts | The signal closest to purchase |
| CTR on SERP | AI Overviews changed the maths | Citation rate when sources are shown | Tells you which content the model trusts |
| "AI visibility score" (vague) | No receipts, no operability | Shortlist share by segment | Tells you exactly where to focus content |
| Backlink count | Diminishing returns for citation | Branded search lift on Amazon | Demand creation, measured downstream |
| Bounce rate | AI traffic behaves differently | PDP conversion rate post-update | Did the optimisation actually pay? |
| "Time on page" | Easy to spoof, hard to act on | Review velocity post-update | Reviews echo the answer — and feed it |
## A worked example: how the swap changes a quarterly review
For an anonymised consumer-brand client running across Amazon and D2C, here's what the new scoreboard looked like at the end of one quarter.
- **Old metric: keyword rank.** Held position 3–5 across 12 head terms. Looked stable. Sales velocity was actually declining 6% quarter on quarter.
- **New metric: Share of Model.** Dropped from 41% to 27% across the prompt set over the same window. Two competitors had launched aggressive content programmes. Our brand was being squeezed out of the AI shortlist while still holding rank.
- **The intervention.** Q&A and bullet rewrites focused on the constraint-led prompts where Share of Model had fallen. Shipped over six weeks. (Methodology: [Lmo7 experiment playbook](/blog/experiment-playbook-feedback-amazon-ai-search-2025).)
- **The recovery.** Share of Model back to 38% by month nine. Sales velocity up 11% on the affected SKUs. Keyword rank didn't move at all.
The old dashboard would have called the quarter "stable." The new dashboard caught the decline 8 weeks before sales did, and gave us the specific lever to pull.
## Bottom line
Retire the metrics that reward noise. Track what AI actually rewards: presence in answers, citations, and recommendation share. Then tie it back to Amazon conversion signals.
That's the new scoreboard.
If you want help building this dashboard for your brand — or auditing the prompt set you should be tracking against — [Lmo7 sets these up every week](/contact).
That is the shift.