Be the product the agent picks.

AI agents now shortlist, compare and buy on your customers' behalf. If your products aren't legible to them, they aren't in the running. We make sure they are.

Free, takes seconds, no sign-up. Checked against ChatGPT, Gemini, Muse, Claude and Perplexity. How it works

Even the best stores leave gaps agents can't read past.

We ran one product page from each of twelve well-known DTC stores through our free check. Eleven could be scored. Read the full teardown.

  • Product schema
  • Price and currency
  • Availability
  • GTIN
  • Ratings in schema
  • Server-rendered text
  • robots.txt access
  • llms.txt

  • Ratings in schema3 readable

    8of 11

    top DTC product pages hide their ratings from AI agents

  • Schema descriptions5 full

    6of 11

    give agents a thin product description, or none

  • All seventeen checks1 clean pass

    1of 11

    passed all seventeen checks

Source: our teardown of twelve DTC stores, one product page each, checked September 21, 2026. Eleven pages could be scored. “Hide” means the rating is not in the structured data the server sends, which is what an agent that doesn't run scripts can read. A chosen sample, not a representative one.

Paste a product URL. See what the agents see.

Instant readiness score and your top three gaps, read from the live page the way an agent reads it. The full report, including what the agents say when asked for the best option in your category, follows by email.

Example readiness report, for illustration only. Your own result opens in a window when you run the check.
Readiness report Example report yourstore.com/products/trail-runner

41/100

Readable in parts. Not yet in the candidate set.

  • No Product schema
  • Stock and delivery missing from feed
  • Returns policy unreadable
  • Checkout passable by agent

Four steps from invisible to recommended.

Take them in order or start where you are. Most clients begin with the audit and have fixes live inside two months.

  1. Audit

    Find every reason an agent skips you.

    • Structured data and product feeds
    • Price, delivery and returns legibility
    Show me more about the audit

  2. Fix

    Make the product machine-readable.

    • Product schema, llms.txt, ACP/UCP feeds
    • Agent-passable cart and checkout
    Show me more about the fix

  3. Cite

    Show up in the sources agents trust.

    • Buying guides, Reddit, review sites
    • Answer-shaped content for full-sentence queries
    Show me more about citations

  4. Track

    Know your share of recommendation.

    • Weekly prompt battery across five agents
    • Named competitors, monthly report
    Show me more about tracking

Stop guessing whether you're on the list.

Before

Not on the list
  • Invisible in agent shortlists
  • No idea what ChatGPT says about you
  • Product data the agent can't parse
  • Checkout that breaks the moment a bot touches it
  • Competitors quietly winning the recommendation

After

On the list
  • In the candidate set for every relevant query
  • A monthly scorecard you can put in a board deck
  • Structured data feeding every major surface
  • A checkout an agent can complete end to end
  • Share of recommendation you can defend and grow

One service. Every surface.

Legibility

  • Product schema
  • Product feeds
  • Variants and stock
  • Delivery and returns

Distribution

  • Agent feeds
  • Merchant Center
  • Marketplaces
  • Buying guides

Trust

  • Reviews
  • Citations
  • Policies
  • Brand mentions

Measurement

  • Prompt battery
  • Candidate set
  • Competitor watch
  • Monthly report

ChatGPT Shopping and Gemini build their answers from product feeds. If your products, prices, stock and policies aren't in a feed they accept, in the format they expect, you are not in the comparison. We build and maintain those feeds over ACP and Google's UCP, and keep them in step with your store.

Why does this exist.

Illustration: an agent scans four products. The two with complete data join the candidate set. The other two are skipped, one for a missing price and one for missing product schema.

Agents don't browse, they filter.

A person will put up with a vague product page and click around until they find what they need. An agent won't. It builds a candidate set from the products it can read, checks each one against the shopper's constraints, and drops anything with a missing answer. There is no second page of results to be found on later. You are either in the set or you are not considered.

Your feed is not your website, and both matter now.

ChatGPT Shopping and Gemini work from structured product feeds. Meta's Muse can open your live store in its own browser and read it the way a customer would. Claude and Perplexity lean on what reviewers and forums say about you. Those are three versions of your product, and they often disagree on price, stock or delivery. When they disagree, the agent trusts none of them.

Nobody gives you a Search Console for ChatGPT.

Search gave you rankings, impressions and clicks. Agents give you nothing: no dashboard, no position, no report of the queries you lost. The only way to know whether you are being recommended is to ask, systematically, every week, with the questions your customers actually use. That is what our prompt battery does.

A catalog is a moving target.

One product page is easy. A catalog is thousands of products with variants, prices that change, stock that runs out, delivery that depends on the ZIP code and a returns policy written by a lawyer. Every one of those has to match across your feed, your markup and your live page, every day. Agents drop the products where they don't. Keeping them in step is the work.

Judge us by the work we publish.

Our first case studies are under way. Until they are out, here is work you can check for yourself.

We asked ChatGPT, Gemini, Claude, Meta’s Muse and Perplexity forty shopping questions and read all 200 answers twice, independently, against a written codebook. The two readings agreed on 99% of 865 brand mentions.

The Agent Index, edition oneRunning shoes under $150, US. Collected 28 September 2026. Method publishedKappa 0.88

The agents disagree. HOKA was recommended in 65% of Gemini’s answers and 13% of ChatGPT’s. Winning one agent is not enough, so we measure all five.

Results by agentSame questions, same day, five agents. 5 agents

Every coded answer is public under CC BY 4.0, so you can check our numbers before you trust them, or use them in your own work.

Open dataResults, coded answers and questions as CSV. 862 coded mentions

Which brands do the agents actually recommend?

We asked five AI agents the same forty questions about running shoes under $150 and counted the brands they told shoppers to buy. Brooks came first. Below the top, the agents barely agree.

See the full Index See the methodology
Running shoes under $150, US. Edition one, September 2026.
RankBrandShareMentioned
1Brooks68%
2ASICS58%
3Saucony56%
4Nike49%
5Adidas44%

Share: answers in which the agent recommended the brand. 200 answers from ChatGPT, Gemini, Claude, Meta's Muse and Perplexity.

Notes from the candidate set.

Start here: how to get your products recommended by ChatGPT and Gemini, and which files AI shopping agents actually read.

All insights

Thirty minutes. One clear next step.

Tell us about your store. You will leave the call knowing where you stand with the agents and what to fix first, whether or not we work together.

What happens next

  1. We reply within one business day with a few times for a call.
  2. We run the check on your store, live, and show you what two agents say about your category.
  3. You leave with your first three fixes, in priority order, in writing.

Rather skip the form?

Pick a time on Calendly (opens in a new tab) hello@recommendedbyagents.com

One partner makes you findable.

Start with the free check, or book a thirty-minute readiness call.

Run the free check Book a call on Calendly (opens in a new tab)

Readiness report

Grade Checked

    Top three gaps

    All checks

    Automated check of one page's public HTML, robots.txt and llms.txt. It measures how readable the page is to AI agents. It doesn't query the agents themselves. The full report does.

    Audit

    Find every reason an agent skips you.

    We test your store the way the agents do: through your feeds, your markup, your live pages and the open web. You get a scored list of what blocks you, in order of what it costs you.

    Timeline
    Two weeks
    Price
    US$950 one-time or US$760 with case-study pricing

    What we look at

    • Product schema, variants, GTINs and identifiers
    • Feed coverage and freshness against your live catalog
    • Price, stock, shipping and returns as an agent reads them
    • Bot walls, modals and scripts that stop a browser agent
    • What five agents say when asked for the best product in your categories

    What you get

    • A readiness score for each template and category
    • A ranked gap list with effort and impact
    • Prompt transcripts and screenshots from each agent
    • A ninety-minute walkthrough with your team
    Run the free check Full details of the Audit service

    Fix

    Make the product machine-readable.

    We close the gaps the audit found, working in your codebase alongside your developers or agency. Nothing ships without their review.

    Timeline
    Four to eight weeks
    Price
    From US$2,900 per project or from US$2,320 with case-study pricing

    What we do

    • Product, Offer, Review and return-policy schema across every template
    • Feeds for ChatGPT over ACP and for Google over UCP and Merchant Center, kept in step with the store
    • llms.txt, and product and policy pages an agent can read without running scripts
    • A cart and checkout an agent can complete: no dead ends, no hostile bot rules
    • Hosted or headless, on the platform you already use

    What you get

    • Pull requests or theme changes, reviewed with your developers
    • Validated feeds with monitoring
    • Before and after readiness scores
    • Documentation your team can maintain
    Run the free check Full details of the Fix service

    Cite

    Show up in the sources agents trust.

    Agents quote what others say about you more readily than what you say about yourself. We find the sources they lean on in your categories and earn your place in them.

    Timeline
    Ongoing first results in six to ten weeks
    Price
    From US$750 per month, with Track or from US$600 per month with case-study pricing

    What we do

    • Map the sources each agent quotes in your categories
    • Earn placement in buying guides, review sites and comparison content
    • Take part where shoppers ask, on Reddit and forums, openly and as the brand
    • Rewrite product and category copy to answer full-sentence questions
    • Build a review base with the volume, recency and detail agents can quote

    What you get

    • A source map for each category
    • A monthly placements report
    • Content briefs and drafts
    • Citation share tracked over time
    Run the free check Full details of the Cite service

    Track

    Know your share of recommendation.

    No agent gives you a dashboard, so we build one. The same questions, asked every week, with your competitors measured alongside you.

    Timeline
    Monthly prompts run weekly
    Price
    From US$750 per month, with Cite or from US$600 per month with case-study pricing

    What we measure

    • Forty or more prompts per category, run weekly across ChatGPT, Gemini, Muse, Claude and Perplexity
    • Candidate-set presence, position and share of recommendation
    • Named competitors tracked alongside you
    • Alerts when a feed breaks or a recommendation drops

    What you get

    • A monthly scorecard you can put in a board deck
    • A live dashboard
    • A quarterly review and plan
    Run the free check Full details of the Track service