How to Do GEO for Ecommerce: A Practical Checklist

If our guide to what AI visibility is answered the why, this answers the how. GEO — generative engine optimization — is the work that gets your store named inside AI answers. Below is the actual checklist, in the order I’d run it for a small brand, using a fictional cat-supplies store, Whisker & Co., to keep every step concrete.

One rule before you start: do these phases in order. There’s no point writing brilliant answer-shaped content if an AI crawler can’t even reach your site, and no point buying a tracking tool to measure a store that hasn’t done the free foundational work. Resist the urge to jump to the shiny parts.

The honest framing. Roughly four-fifths of what follows is free. This guide names tools where they help, and some links may be affiliate links, but the core work here costs effort, not money — and we’ll flag every point where a paid tool is optional rather than necessary.

Phase 1: Make sure engines can read you

This is the least glamorous phase and the one that quietly sinks the most stores. If an AI crawler can’t fetch your pages, nothing else on this list matters — you’re invisible by default. And here’s the uncomfortable part: surveys in 2026 found that roughly a quarter to a third of ecommerce and SaaS sites are accidentally blocking major AI crawlers, usually at the CDN layer, often as leftover settings from the 2023 “block the scrapers” panic.

• Confirm AI crawlers aren’t blocked in robots.txt.Each engine runs separate bots for training vs. live search. Blocking the search bot is what kills your citations.
• Check your CDN / WAF isn’t overriding the file.Cloudflare’s “Block AI bots” toggle silently overrides robots.txt. This is the single most common cause of accidental invisibility.
• Make sure content renders server-side.If your product copy only appears after JavaScript runs, many crawlers see an empty page.
• Don’t noindex pages you want cited, and keep your sitemap allowed.noindex hides a page from AI search bots too. If it should be cited, it must be indexable.

For a store that wants to be found, the right default is simple: allow the search and user-facing bots. Opting out of model training is a separate, optional IP decision that doesn’t affect whether you’re citable. Here’s a clean, visibility-first starting point:

# robots.txt — let AI answer engines read & cite you
# OpenAI (ChatGPT)
User-agent: OAI-SearchBot   # powers ChatGPT search citations
Allow: /
User-agent: ChatGPT-User    # user-triggered fetches
Allow: /
# Anthropic (Claude)
User-agent: Claude-SearchBot
Allow: /
User-agent: Claude-User
Allow: /
# Perplexity, Google, Bing, Amazon, Apple
User-agent: PerplexityBot
Allow: /
User-agent: Googlebot        # never block this one
Allow: /
User-agent: Google-Extended  # Gemini / AI Overviews
Allow: /
User-agent: Bingbot
Allow: /
User-agent: Amazonbot        # Amazon Rufus
Allow: /

# Optional — opt OUT of model *training* only (an IP choice;
# does NOT affect the search bots above):
# User-agent: GPTBot
# Disallow: /
# User-agent: ClaudeBot
# Disallow: /

Sitemap: https://yourstore.com/sitemap.xml

Whisker & Co. ran this check and discovered its Cloudflare bot rules were quietly turning away two retrieval crawlers. A ten-minute fix unlocked everything downstream — the highest-leverage thing it did all month.

{%- for group in robots.default_groups -%}
{{ group.user_agent }}
{%- for rule in group.rules -%}{{ rule }}{%- endfor -%}
{%- if group.sitemap != blank -%}{{ group.sitemap }}{%- endif -%}
{%- endfor -%}

User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: Claude-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /

Phase 2: Make your catalog machine-readable

Engines lean heavily on structured data — schema markup — to understand exactly what you sell. Without it, a crawler is squinting at your page guessing which number is the price. With it, you’re handing the model clean, labeled facts it can quote confidently. This phase is free and high-leverage; most platforms support it with an app or a template edit.

• Add Product schema to every product page.Name, brand, SKU/GTIN, description, image, and an Offer with price and availability.
• Add AggregateRating + Review.Star rating, review count, and real review text — a primary trust signal for engines.
• Describe attributes with additionalProperty.Noise level, size, capacity, “suits 2 cats” — the exact fields engines quote for narrow questions.
• Add FAQPage schema to product and category pages.Answer-shaped Q&A blocks are disproportionately easy for engines to lift.
• Add Organization schema with sameAs.Link your social and marketplace profiles so the model treats you as one entity.

Here’s the shape of the core markup — a trimmed Product + rating example to adapt. Wrap it in a <script type="application/ld+json"> tag in your page head:

{
"@context": "https://schema.org",
"@type": "Product",
"name": "Quiet Auto Litter Box",
"brand": { "@type": "Brand", "name": "Whisker & Co." },
"sku": "WC-LB-01",
"description": "Self-cleaning litter box for small apartments; quiet enough for a studio, handles up to two cats.",
"additionalProperty": [
  { "@type":"PropertyValue", "name":"Noise level", "value":"38 dB" },
  { "@type":"PropertyValue", "name":"Capacity", "value":"Up to 2 cats" }
],
"offers": { "@type":"Offer", "price":"249.00", "priceCurrency":"USD", "availability":"https://schema.org/InStock" },
"aggregateRating": { "@type":"AggregateRating", "ratingValue":"4.6", "reviewCount":"218" }
}

Validate it with Google’s Rich Results Test or Schema.org’s validator before shipping. A small note that matters for stores: research suggests a large share of ecommerce sites implement product schema incorrectly, so getting this right is itself an edge.

Phase 3: Write answer-shaped content

Most product and blog copy is written to rank for keywords. Engines don’t want keywords; they want the answer. The shift is simple but profound: stop writing “Automatic Litter Boxes | Whisker & Co.” and start writing the page that directly answers a real question a shopper would ask an assistant — then back it with honest detail.

• Lead with the answer.Open the page with a direct, one-paragraph answer to its core question. Don’t bury it six headings down.
• Say who it’s for and what it solves, in plain language.“Quiet enough for a studio; handles two cats” beats a spec table the model can’t map to intent.
• Cover the four content types that match how people ask.Use-case, attribute, comparison, and branded pages — mirror the way shoppers actually prompt.
• Build genuine comparison and round-up pages.Engines lean on comparative content; an honest “best X for Y” page is quotable gold.

Phase 4: Earn trust where engines listen

This is the hard, slow, decisive phase — and the one no tool can do for you. Engines trust other people’s pages about you more than your own. Analyses of large samples of AI citations keep finding third-party sources — Reddit especially — dominating what gets quoted for shopping questions. Which means the highest-value GEO work happens off your own site.

• Be genuinely useful in the communities your buyers use.Real participation in the relevant subreddits and forums — not astroturf, which engines are tuned to discount.
• Earn honest reviews on independent sites and round-ups.A spot in a credible “best litter boxes” list does more than fifty of your own blog posts.
• Get attribute-rich reviews on your own pages.Nudge customers to mention specifics — apartment size, number of cats, noise — the long-tail evidence engines quote.
• Encourage video and unboxings.YouTube is a frequently-cited source; one honest demo compounds.
• Be one consistent entity everywhere.Same name, description, and key facts across your site, socials, marketplaces, and directories.

Phase 5: Measure & maintain

You can’t improve what you don’t track, and you shouldn’t buy tooling before you’ve felt the manual version. Establish a baseline, watch it monthly, and aim each round of work at the queries you’re still losing.

• Build a prompt library of 20–50 real questions.Across the four content types above. Use the four prompt buckets above as a ready-made starting set.
• Run it monthly across ChatGPT, Perplexity, and Google AI Overviews.Log three things per answer: were you named, described accurately, and where versus competitors.
• Track your AI Visibility Score, sentiment, and owned-vs-earned share.Named in 6 of 30 prompts is a 20% score. Track competitors the same way.
• Keep the feed and reviews fresh.Accurate stock and prices, and a steady trickle of new reviews. Freshness is a trust signal.
• Re-aim at lost queries, and re-check phase 1 after site changes.A CMS upgrade can silently re-block crawlers — re-test periodically.

Start this by hand; it costs only attention. When tracking enough prompts across enough engines genuinely hurts, that’s the signal to look at a tool — see our vendor-neutral tool comparison, where the honest first recommendation for most stores is free.

Recap: The one-page GEO checklist

Here’s the whole thing in one scannable list. Tap an item to check it off as you go.

Pitfalls: What not to do

Three mistakes account for most wasted GEO effort. Skipping phase 1 — doing beautiful content work while a CDN quietly blocks the crawlers — is the most common and the most painful. Astroturfing phase 4 — flooding Reddit with thinly veiled promotion — backfires, because engines and communities both discount obvious manipulation. And buying tools before doing the free work — paying for a dashboard to measure a store that hasn’t fixed its schema or crawlers — is spending money to watch a problem you haven’t started solving. Do the boring, free things first; they’re where the leverage is.