What an Ecommerce AEO Agency Does in the First 30 Days
A plain account of VendaHawk's first month with a client: the snapshot, onboarding and access, the brand guide, the baseline prompt panel, the first fixes, approvals and measurement, plus what we cannot promise.
- By
- VendaHawk
- Published
- Updated
- Reading time
- 7 min read
In the first 30 days, VendaHawk takes a snapshot of how AI assistants describe your products today, collects access without ever asking for a password, writes a profile of your brand that everyone on our team follows, runs a baseline panel of shopper prompts across the engines, ships the first approved fixes, and sets up weekly measurement. You see every draft before it goes live. We do not promise rankings, because nobody can.
How the work gets done
VendaHawk is a small team, and each piece of work has one owner: Amazon bullets, A+ modules, Product markup for a brand site. Everyone works from the same written brand guide, every draft is checked against the platform's rules and reviewed by a second person, and anything that touches your customers waits for your approval.
Week 1: snapshot, onboarding and access
The snapshot. Before anything else, we ask the engines about your category. On the discovery call we query ChatGPT, Google AI Mode, Perplexity and Amazon's assistant on three category prompts, record who is cited and whether you appear, and agree on five questions you want to own.
The onboarding form. The best agencies collect four things, in this order: access, baselines, constraints, goals. Our form follows that order across seven steps and saves progress as you go. It asks for your business in your own words, hero products with price and margin, your ad-efficiency ceilings if we manage ads, three to five competitors, the claims that need evidence, and who approves what.
Access without passwords. Every platform has a delegated route, and we only use those:
| Platform | What you do | Role we ask for |
|---|---|---|
| Amazon Seller Central | Settings > User Permissions > invite our email | View and Edit on inventory, A+, advertising, reports; never account or bank settings |
| Amazon Ads | Share your Account ID; approve our manager request | Editor |
| Walmart Seller Center | People and permissions > Add User | Read and Write |
| eBay | My eBay > Account > Team access | Listings, orders, promotions, advertising |
| Shopify | Settings > Users > Security > send the 4-digit collaborator code | Products, themes, marketing, analytics as needed |
| Google Merchant Center, GA4, Search Console | Add our email with a role | Standard, Marketer, Full |
| Meta and TikTok | Add our business portfolio or Business Center ID as a partner | Manage campaigns, create content |
The kickoff is gated on access being complete, because baselines are pulled by API as soon as roles land.
Week 2: the brand guide and the baseline panel
The brand guide. About 25 answers from the form and the discovery call become a written document our team follows: voice sliders and three samples of copy that sound like you, a do-not-say list, claims that need substantiation and the evidence for each, product fact sources per hero SKU (spec sheet, certifications, materials, dimensions, warranty), margin floors, dated goals, approvers and turnaround times. The form wins on numbers, the call wins on tone, and conflicts come back to you. The guide is versioned, and we update it whenever an answer changes.
The baseline prompt panel. We build about 50 prompts from your product titles, Amazon Q&A, reviews and Merchant Center's top terms, spread across discovery, evaluation and ready-to-buy stages. We run them on ChatGPT, Perplexity, Google AI Mode and Copilot, several samples each, and record share of voice against your competitor set as the baseline for every later report.
Technical readiness. In the same week we check robots.txt and CDN bot rules against the search crawlers that matter (OAI-SearchBot, PerplexityBot, Googlebot, Claude-SearchBot, Amzn-SearchBot, bingbot), validate Product markup for duplicates, confirm Shopify's Agentic Storefronts channel is active, pull Merchant Center's AI performance report and its missing-attributes list, and audit Amazon listing completeness, Walmart Listing Quality and eBay item specifics coverage.
Week 3: first fixes and approvals
For each hero SKU we draft an answer-first paragraph, a spec table with units, a FAQ built from real questions, and a comparison block, then map the same facts to Amazon bullets and A+, Walmart key features, eBay item specifics and description HTML, and the brand site. Numbers stay identical everywhere, because conflicts between fields are the most common cause of wrong AI answers.
These always wait for a named approver:
- Any change to price, availability, shipping or return policy
- Any claim about safety, certification, compatibility or performance
- Brand voice on customer-facing pages
- Opting a channel in or out: Agentic Storefronts, Copilot Checkout, Google's UCP checkout, Perplexity's Merchant Program
- Your policy on training crawlers such as
GPTBotandClaudeBot, decided once
Your approval matrix sets who signs each asset type and whether an item auto-approves after 24, 48 or 72 hours of silence, or never.
Week 4: publish, measure, report
Approved changes go live through the platform tools or APIs, feeds are resubmitted, recrawls are requested, and any programme you approved gets its application sent. Then the prompt panel runs again. The first report shows share of voice per engine against baseline, the pages engines cited instead of you (sorted into editorial lists, YouTube, Reddit, review platforms and retailers), Merchant Center deltas, Walmart Listing Quality movement, and AI referrals in GA4. On Shopify, the thank-you page survey with "An AI assistant" as an answer option starts collecting first-touch data.
From that report comes the next backlog: missing attributes, FAQ gaps from new prompts, pages older than 12 months, price or availability conflicts, and crawler-block regressions. Month two starts with outreach to the third-party pages engines cite.
What you have to do
- Complete the form and grant access in week 1; nothing moves until roles are in place.
- Attend one 45 to 60 minute discovery call once the form is at least 70% complete.
- Send spec sheets, certifications and three copy samples. We never invent product facts; a missing fact becomes a question back to you.
- Name an approver per asset type and answer approvals within your chosen window.
Honest limits
- No guaranteed rankings or citations. AI answers vary from run to run, which is why we sample repeatedly and report rates, not positions.
- Brand-owned sites are a small share of citations in AI shopping answers, between roughly 3% and 6% across engines in AEOsome's 2026 study (AEOsome). Much of the work is earning mentions on pages you do not own.
- Being cited is not being recommended. Ahrefs found 43% of answers that cited a promotional page never named the brand (Ahrefs).
- Amazon publishes no field weights for its assistant and Walmart publishes no ranking inputs for Sparky; we optimise what each platform documents.
- OpenAI's direct feed is still limited to approved partners, so most stores reach ChatGPT through Shopify or the open web.
- An llms.txt file is cheap and we ship it, but no engine officially uses it and we do not report it as a result.
- API-sampled answers differ from what a shopper sees in the app. We say which we measured.
Frequently asked questions
Will I be asked for any passwords?
No. Every platform we work in has a delegated access route: secondary users on Seller Central, manager links on Amazon Ads, Team access on eBay, collaborator codes on Shopify, roles on Google and partner assignment on Meta and TikTok. We keep a register of what was granted and when to revoke it.
What is a brand guide?
A versioned written profile of your brand that everyone on our team follows: voice and samples, banned words, claims and their evidence, product facts per SKU, margin floors, goals, approvers and turnaround times. It is built from your onboarding answers and the discovery call, and we update it whenever you change an answer.
What do I see before something goes live?
A draft and a diff, in the portal, with the rule checks it passed. Price, availability, policy, claims, brand voice and channel opt-ins always wait for a named approver. Other items can auto-approve after a window you set, or never.
How do you measure whether it worked?
A fixed prompt panel run on a schedule across ChatGPT, Google AI Mode, Perplexity and Copilot, scored for mentions, position and citations against your competitor set, plus Google's own Merchant Center AI performance report, AI referrals in analytics and the thank-you page survey on Shopify.
Can you guarantee my product shows up in ChatGPT?
No, and you should be wary of anyone who does. Answers are generated, vary between runs, and draw mostly on third-party pages. We can show baseline against current share of voice per engine, the exact fixes shipped, and which pages engines cite instead of you.
Sources
- https://developers.google.com/search/docs/appearance/ai-features
- https://support.google.com/merchants/answer/17200695
- https://developers.openai.com/api/docs/bots
- https://developers.openai.com/commerce/guides/get-started.md
- https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts
- https://www.amazon.science/blog/the-technology-behind-amazons-genai-powered-shopping-assistant-rufus
- https://www.aeosome.com/blog/research/ai-shopping-visibility-study-2026/
- https://ahrefs.com/blog/self-promotional-content-ai-seo-experiment/
- https://shopify.dev/docs/apps/build/dev-dashboard/stores/collaborations
- https://www.ebay.com/help/account/multiuser-account-access/multiuser-account-access?id=5197
- https://www.leadsie.com/blog/give-amazon-ads-access
- https://shopify.dev/docs/apps/build/checkout/thank-you-order-status/add-survey
- https://llmstxt.org/
Questions this article answers
Will I be asked for any passwords?
No. Every platform we work in has a delegated access route: secondary users on Seller Central, manager links on Amazon Ads, Team access on eBay, collaborator codes on Shopify, roles on Google and partner assignment on Meta and TikTok. We keep a register of what was granted and when to revoke it.
What is a brand guide?
A versioned written profile of your brand that everyone on our team follows: voice and samples, banned words, claims and their evidence, product facts per SKU, margin floors, goals, approvers and turnaround times. It is built from your onboarding answers and the discovery call, and we update it whenever you change an answer.
What do I see before something goes live?
A draft and a diff, in the portal, with the rule checks it passed. Price, availability, policy, claims, brand voice and channel opt-ins always wait for a named approver. Other items can auto-approve after a window you set, or never.
How do you measure whether it worked?
A fixed prompt panel run on a schedule across ChatGPT, Google AI Mode, Perplexity and Copilot, scored for mentions, position and citations against your competitor set, plus Google's own Merchant Center AI performance report, AI referrals in analytics and the thank-you page survey on Shopify.
Can you guarantee my product shows up in ChatGPT?
No, and you should be wary of anyone who does. Answers are generated, vary between runs, and draw mostly on third-party pages. We can show baseline against current share of voice per engine, the exact fixes shipped, and which pages engines cite instead of you.