Your product got mentioned in an AI answer. Did anyone buy anything? Most AI visibility dashboards stop at the mention. An ecommerce GEO tool has to carry the story from the answer a shopper reads to the order in your store. It also has to show how much of your buying journey it actually measured.
- A share-of-voice chart is not a revenue report. Mentions tell you an assistant said your name. AI-attributed revenue estimates what that answer moved, with the method behind the estimate on screen.
- A conversion score without coverage is a rumor. The same score over 3% of answers and over 90% of answers rest on very different amounts of evidence.
- Absence has to read as absence. When a brand does not appear, Arenza records AVS 0 and does not score ACS at all. A missing brand never looks like a badly-selling one.
Quick verdict
What Arenza gives an ecommerce team, and the lowest tier that carries it. List prices only.
| What you want | What Arenza gives you | Lowest tier that carries it |
|---|---|---|
| See how AI answers describe you | Verbatim answers from ChatGPT, Google AI Mode and Perplexity | Free, $0, no credit card |
| A visibility and a conversion score you can read | AVS and ACS, with ACS always shown alongside its coverage | Free |
| Catch wrong claims about your catalogue | Outdated specs, false claims, brand-attribution errors and miscategorization, each with severity, frequency and the verbatim quote | Free |
| Cover the whole buying journey | All buyer intents scanned, weekly, in one market | Free |
| Compare against the brands you actually lose to | Unlimited tracked competitors with head-to-head comparison | Starter, $49/month |
| AI-attributed revenue, once your order data is connected | API and MCP so conversion events reach Arenza | Starter, $49/month |
| Get the work executed, not only reported | Articles hosted on your own domain, on-site technical fixes deployed for you, 3 markets, 3 brands | Pro, $299/month |
| One screen for a roster of 10 to 200 client brands | Portfolio view, CSV bulk import, copy-settings A→B, white-label | Performance, outcome-based and sales-led |
1. A mention is an impression, and ecommerce has never been paid for impressions
Ecommerce teams already know how to read a funnel. Sessions, add-to-carts, orders, revenue. Then AI search arrived and the reporting reverted to exposure vocabulary: mentions, share of voice, prompt counts.
Share of voice is a legitimate media metric. It just does not answer the question a merchandiser is asked in a Monday meeting. Did this channel produce orders?
Arenza is built around three headline metrics: AVS, ACS and AI-attributed revenue. They are read together. Visibility and conversion explain the movement; revenue estimates the outcome.
2. AVS, ACS and revenue are read together, and reading them out of order misleads you
Think of it as three linked questions rather than three unrelated charts.
AVS — were you found? The Agentic Visibility Score measures whether ChatGPT, Google AI Mode and Perplexity surface your brand for the buying questions your customers ask. No presence, nothing downstream. A prompt that already names your brand does not count as visibility earned, so that case is marked N/A rather than scored.
ACS — once found, did the answer sell you? The Agentic Conversion Score measures whether the answer actually sells you once you are found. A weak ACS with a strong AVS is a specific, fixable problem. You are in the conversation and losing it.
AI-attributed revenue — did it turn into orders? This is the estimate an ecommerce team takes to finance, and section 5 explains how it is built.
The order matters because the middle number is conditional. ACS is scored only over the answers where your business actually appeared. That is the difference between a score describing your persuasion and a score contaminated by absence.
3. Absent is not the same as unpersuasive, so absence must never score zero
Here is the trap that ruins AI dashboards for ecommerce.
Suppose an assistant answers a hundred shopper questions in your category and never names you once. A naive scoring system marks a zero for conversion. That zero reads as a claim: the AI discussed your brand and failed to sell it. That claim is false, and it points your team at the wrong repair.
Arenza treats absence as a visibility fact. When a brand does not appear, AVS is 0 and ACS is not scored at all. Never 0.
The practical effect is that your two numbers stay honest about different failures. AVS 0 sends you to reach work: coverage, sources, the pages assistants read. A low ACS over real coverage sends you to persuasion work: what the answer says about your price, your fit, your specs.
ACS is always shown with its coverage. A high score measured over three answers rests on thin evidence. The same score over most of the answers in your category rests on broad evidence.
4. Ecommerce also needs the answer to be correct, not just present and warm
Presence and persuasion still assume the assistant is describing the right product. Often it is not.
Assistants repeat a discontinued spec. They attribute a rival's feature to you. They put you in the wrong category, which removes you from every shortlist that category feeds. For a catalogue business, a wrong spec is a return, a support ticket, and a refund.
Arenza treats accuracy as a diagnostic surface rather than a headline score. It detects outdated specs, false claims, brand-attribution errors and category miscategorization. Each finding carries severity, frequency and the verbatim quote from the answer.
The verbatim quote is the working part. You can hand the exact sentence to a product manager. The argument then moves to the useful question: is the model reading your page, a stale cache, or a third-party review?
5. The revenue number is an estimate, so you should be able to see how it was built
A vendor that reports your AI revenue is reporting on its own work. That deserves a method, not a confident round number.
Arenza observes the parts that are observable. A session arrives, and a landing page is recorded. The link from that landing page back to the prompt a shopper typed is inferred, weighted by which sources the answers cited. It is an attributed estimate, not a direct observation of an answer causing an order.
Two things follow. Every attributed event keeps an audit trail you can open, so the number is checkable rather than declared. And Performance, the outcome-based tier, is designed to settle on high-confidence server-side events rather than on soft inference.
Getting there needs order data. Conversion events reach Arenza through the API and MCP server, which start on Starter at $49/month. Free shows you the answers, AVS, ACS and accuracy without any order connection.
When the numbers need to leave the dashboard, there is an MCP server at https://api.arenza.ai/mcp with OAuth 2.1 and dynamic client registration. A JSON-RPC adapter at https://api.arenza.ai/rpc is gated by a personal access token. Alerts route per brand to Slack, Feishu or a webhook. The interface is bilingual in zh-CN and en-US.
Arenza measures the US, UK, DE and JP markets, but no single purchasable tier gives you all four. Free and Starter cover one market, and Pro covers three. Brands work the same way: Free and Starter cover one, Pro covers three, and a 10-to-200-brand portfolio is the Performance and agency arrangement.
Cost math for an ecommerce team
List prices only, with the billing basis, because the basis changes the real commitment. Confirm each vendor's current pricing on their own page before you buy.
| Option | List price | Billing basis | Published engines and limits |
|---|---|---|---|
| Arenza Free | $0 | No credit card | 3 AI assistants, AVS + ACS, all buyer intents, weekly scan, 1 market, verbatim answers, brand accuracy, cited brands as evidence, ARC Autopilot |
| Arenza Starter | $49/month | Cancel anytime | Daily scans on 3 assistants across all intents, 1 market, unlimited tracked competitors, 5 articles/month, API and MCP, Slack and Lark |
| Arenza Pro | $299/month | Cancel anytime | All three AI assistants scanned daily, 3 markets, 3 brands, unlimited competitors, 100 articles/month, articles on your own domain, on-site technical fixes deployed for you, off-site analysis, Arenza Agents, white-label |
| Profound Starter | $99/month | Billed yearly, 2 months free | Published as ChatGPT tracking only; 50 prompts tracked, 1,500 responses monthly, 1 language, 1 region |
| Profound Growth | $399/month | Billed yearly | 3 answer engines: ChatGPT, Perplexity, Google AI Overviews; a "Try for free" option is published on this plan |
| Profound Enterprise | Custom | Not published | Up to 9 answer engines |
| Otterly.AI Lite | $29/month | Monthly | 15 search prompts |
| Otterly.AI Standard | $189/month | Monthly | 100 prompts; API and MCP access published on this plan |
| Otterly.AI Premium | $489/month | Monthly | 400 prompts |
| Otterly.AI Enterprise | From $1,000/month | Custom | Core engines ChatGPT, Google AI Overviews, Perplexity, MS Copilot; Claude, Gemini and Google AI Mode are paid add-ons |
| Ahrefs Brand Radar, Custom Prompts | From $50/month, on top of a paid Ahrefs base plan | Included in Ahrefs paid plans Lite and above | AI Overviews, Gemini, Perplexity, ChatGPT, Copilot, AI Mode; Claude as add-on |
| Ahrefs AI Visibility Index | $199/month | Not published | Same published engine coverage as above |
Two things fall out of the table. A yearly billing basis is a yearly commitment, whatever the monthly figure reads. And the Ahrefs figures sit on top of an existing paid Ahrefs plan, so $50 is not a standalone total.
Arenza's Free tier is $0 with no credit card, so a merchandiser can scan the buying journey before anyone signs anything. Performance is outcome-based and sales-led at hello@arenza.ai. Arenza runs ARC — Answer, Reach, Convert — end to end and is paid when the agreed outcome happens. Each success-fee structure is agreed per customer in the contract.
Arenza meters tracked prompts too, and the number belongs next to everyone else's. Arenza tracks 10 prompts on Free, 30 on Starter and 120 on Pro. Buyer intents are uncapped on every tier, but tracked prompts are not, so compare prompt counts against prompt counts.
The choice comes down to this
If you only need to know whether ChatGPT says your name, start on Free at $0 and read the answers yourself.
If you are trying to explain an AI channel to a CFO, start free to establish AVS and ACS. Then connect your order data on Starter, where API and MCP make the AI-attributed revenue estimate possible.
If your problem is that AI answers describe your catalogue incorrectly, prioritise accuracy diagnostics with severity, frequency and the verbatim quote. Those are on Free.
If you sell in more than one market, note the per-tier limit before you compare prices. Free and Starter cover one market, and Pro covers three.
If you carry a roster of client brands rather than one, the portfolio view and bulk import live in the agency and Performance arrangement, so talk to sales.
FAQ
How do I know if an AI assistant actually drove a sale?
You get an attributed estimate rather than a direct observation, so ask any vendor to show its method. Arenza observes the session and the landing page, then infers the link back to the prompt using the sources the answers cited. Each attributed event keeps an audit trail you can open. Arenza reports AVS, ACS and AI-attributed revenue together so visibility, persuasion and outcome are read as one story.
Why does a conversion score need to show coverage?
Because the same score over a small sample and a large sample rest on very different evidence. A strong score measured over 3% of answers is thin evidence. The same score over 90% of answers is broad evidence. Arenza always shows ACS with its coverage for that reason.
What happens to my scores if AI answers never mention my brand?
Your visibility score is 0 and your conversion score is not calculated at all. A zero conversion score would wrongly imply the assistant discussed your brand and failed to sell it. Keeping it unscored points you at the real problem, which is reach.
Can I test this before paying anything?
Yes. Arenza has a Free tier at $0 with no credit card, covering three AI assistants, one market, weekly scans and all buyer intents. You see verbatim AI answers, AVS, ACS and brand accuracy findings on that tier. Revenue attribution is not part of Free, because it needs your order data connected through the API on a paid tier.
Which AI assistants and markets should an ecommerce brand measure?
Measure the assistants your buyers actually use and the markets you actually ship to. Arenza measures ChatGPT, Google AI Mode and Perplexity across the US, UK, DE and JP. No single tier covers all four markets: Free and Starter cover one, and Pro covers three.