Adam Clay

In a Q&A post drawn from hundreds of conversations with brands and retailers, the answers reach the same conclusion: what isn't encoded in structured form effectively doesn't exist to an agent. Content teams, then, need to fix product-level data before pursuing AI visibility—a sequencing Clay presents as a brand's main decision.

Product images are invisible to AI agents, Clay writes, because reading an image costs far more than reading text.

• Specifications, sizing and materials shown only in pictures must also appear in metadata and schema.
• They explain why the same product looks different across agentic platforms: data flows from a PIM to the brand site, through separate feeds to each retailer and from another system to direct-to-consumer sales, with manual steps in between.

Read original · Original 2

Agents use it to establish what a product is, then look for where to buy it—reversing the longstanding assumption that a brand site is an afterthought. A thin, unstructured brand page can therefore carry a direct commercial cost.

For brands that have treated their own sites as an afterthought, the point is clear: if brand-dot-com is thin or unstructured, a product may never enter an agent's consideration set, however strong a retailer's listing.

Read original

They credit its real uses—site-level crawlability, sentiment monitoring and brand PR inputs—while rejecting claims that AI visibility dashboards are actionable, repeatable and tied to business outcomes.

• Prompts and answers shift with user context and over time, they argue, so treating AI prominence like an SEO rank is the wrong mental model.
• Their takeaway: buy these tools for marketing, PR and crawl hygiene, but don't confuse visibility with growth.

Clay names opportunity cost as the biggest risk: teams spend six to eighteen months chasing dashboards instead of fixing product-level content that is measurable and drives sales.

• The post also flags black-hat tactics highlighted by the Wall Street Journal, questionable panel data behind prompt prediction and shrinking citations as transactions move upstream.
• Clay credits much of the assessment to a Retailgentic piece by Scot Wingo, says some GEO/AEO work has value and invites disagreement.

Read original

They pair conversion arguments with a feature timeline, framing platform competition as a "battle of fiefdoms." These are circulated predictions, not measured results; Clay attributes the substance to Retailgentic and Scot Wingo.

The question-and-answer post states that agentic traffic converts better than organic because a shopper told by an assistant to buy a specific product generally buys where the assistant points, and that agentic traffic is predicted to pass traditional search in under two years. A separate post relays Retailgentic reporting that a ChatGPT universal shopping cart is imminent, quotes Scot Wingo's view that ten to twenty companies have tried a cross-merchant cart since 1999 and all failed, and repeats Wingo's prediction that seven typical years of agentic-commerce features will be packed into 2026, producing an "epic Holiday 26."

Read original · Original 2

They consistently name the source, an important distinction when assessing their own position: the underlying analysis is usually someone else's. Ask for the deep dives behind the posts to assess that analysis.

One post describes a Retailgentic newsletter as full of news, gossip and analysis, highlighting a first look at ChatGPT instant checkout, the possibility of Reddit shopping and candid takeaways from BigCommerce and Shopify earnings calls.

• The full newsletter is behind a comment link.
• Another offers only the framing text for a two-part deep dive into Shopify's agentic catalog and eleven Shopify questions answered by Wingo.

Clay amplifies Wingo's work, while Wingo's GEO/AEO analysis supplies the argument Clay adopts and restates in their own takeaway. The Shopify part-two post provides only a framing description; ask for the underlying analysis before assessing their position on it.

Read original · Original 2 · Original 3 · Original 4

If you read one item, read this one.

Adam Clay frames this as answers drawn from "hundreds of conversations with the biggest brands and retailers," written as five question-and-answer pairs.

• The answers carry specific positions: product images are effectively invisible to AI agents because reading images costs more than reading text, so anything shown only in a picture — specs, sizing, materials — has to also appear in metadata and structured data.
• Agentic traffic is still a small share of site visits but converts better than organic, because a shopper told by an assistant to buy a particular product tends to buy it where the assistant points.
• Different platforms represent the same brand differently because product data comes from disconnected systems: a PIM feeding the brand site, separate feeds for each retailer, another system for direct-to-consumer, and manual steps in between.
• On measurement, Clay says the link between content work and agentic-commerce return will only become obvious as platforms open up more data, and recommends starting with visibility — which products appear, which do not, which are outranked by resellers.
• On the role of a brand's own website, the post argues the brand domain becomes the canonical definition of what a product is, so a thin or unstructured brand page can keep a product out of consideration regardless of how good a retailer's listing is.
• The post closes by naming ReFiBuy.ai as a partner that can help.

For brands planning a campaign

The five questions map almost exactly onto the objections brand teams raise in early agentic-commerce conversations, making them a natural spine for a B2B ecommerce platform's content collaboration.

• Each answer could be extended with the platform's own documentation, schema guidance, or platform-by-platform notes.
• Two practical conditions: the closing promotion of ReFiBuy.ai means a partner would need to agree in advance how their product is named alongside it, and any collaboration should keep the measurement caution intact rather than promise dashboard-level ROI.

This post assesses generative-engine and answer-engine optimization tools, mostly summarizing a Retailgentic piece with Scot Wingo.

• The good: the tools are trending toward free, they can drive real crawlability improvements at the site level, and they are decent for sentiment monitoring and brand PR inputs.
• The bad: AI visibility dashboards often are not actionable, repeatable, or tied to business outcomes; prompts and answers vary by user context and time, so "AI conversations are snowflakes" and measuring rank the way SEO rank is measured is the wrong mental model; and citations are shrinking as search drifts toward zero-click, with transactions moving upstream.
• The ugly: some so-called best practices veer into manipulation, which the post says has been highlighted by the Wall Street Journal, and it names First Page Sage; panel data behind prompt prediction is questionable; and the biggest risk is opportunity cost when teams chase dashboards for six to eighteen months instead of fixing product-level content that is measurable and drives sales.
• Clay's stated takeaway is to use GEO and AEO tools for marketing, PR, and crawl hygiene, but not to confuse visibility with a growth strategy — the real work is product-level storytelling and structured data that language models can read consistently.
• The post invites disagreement.

For brands planning a campaign

A platform selling agentic-commerce readiness can use this framing to argue buyers should invest in product data quality rather than visibility dashboards. Because most of the argument is relayed from Retailgentic, any collaboration reusing it should credit that source rather than present the analysis as Clay's own data.

Clay relays Retailgentic reporting that a ChatGPT universal shopping cart feature is imminent.

• The post quotes Scot Wingo saying he has been in ecommerce since 1999 and has watched ten to twenty startups and large companies attempt a universal, merchant-friendly cart that works across more than five merchants, with all of them failing.
• Wingo predicts ChatGPT and others will pack "seven typical years" of agentic-commerce features into 2026, with cart functionality playing a large part, and calls for an "epic Holiday 26."
• Clay frames the news as a major shift and points readers to Wingo and Retailgentic for breaking coverage.

For brands planning a campaign

This forward-looking item is useful for anyone building a 2026 calendar around agentic commerce, and shows where Clay expects attention to concentrate. It also casts them as curator and amplifier of Wingo's reporting rather than original analyst, so a collaboration that depends on proprietary data or forecasts would need to bring that data from elsewhere.

A brief pointer post recommending a Retailgentic newsletter described as full of news, gossip, and analysis from the worlds of ChatGPT, Shopify, and BigCommerce.

• Clay highlights a first look at ChatGPT instant checkout features, the possibility of Reddit shopping, and candid takeaways from the BigCommerce and Shopify earnings calls, and says the newsletter is worth reading in full.
• The post directs readers to a link in the comments.

For brands planning a campaign

Useful mainly as a register signal: Clay mixes practical product-data advice with personality-driven news commentary.

• A brand collaboration can lean into the conversational tone, but the substantive claims live in the linked newsletter.
• Ask to review it before quoting those claims.

The post describes a two-part Retailgentic deep dive into Shopify's agentic catalog, plus eleven Shopify questions and answers with Scot Wingo. Ask for the original deep dive before making claims about how Shopify's catalog works.

For brands planning a campaign

Acts as a pointer to a Shopify-specific reference. Any collaboration making claims about Shopify catalog specifics should verify them against the original piece; ask for the full text before releasing the page.

The most detailed example is a five-question breakdown drawn from conversations with brands and retailers.

• Product images are effectively invisible to agents because they cost more to process than text, so anything shown in an image also needs to appear in metadata and schema.
• Brand websites serve as the canonical description of a product even when most sales run through retailers; a thin or unstructured brand site can keep a product out of an agent’s consideration set.
• They also note that product data often flows from a PIM, separate retailer feeds and a different DTC system, leading the same product to be represented differently across agentic platforms.

A Q&A distilling recurring questions from hundreds of conversations with brands and retailers, alongside commentary on Retailgentic analysis.

The five-question post closes with a recommendation for ReFiBuy.ai, a commercial service; separate that promotion from the analysis.

For brands planning a campaign

A platform collaboration could focus on practical product-data audits or metadata and schema guidance, where Clay’s argument is most specific and clearly tied to outcomes.

Examples: Source 1 · Source 2 · Source 3

They can improve site-level crawlability and support LLM-era sentiment monitoring and brand PR.

• But AI visibility dashboards are often not actionable, repeatable or tied to business outcomes.
• Prompts and answers vary by user context and over time, making rank tracking the wrong model, while citations shrink as transactions move upstream.
• They also flag black-hat tactics and call opportunity cost the biggest risk: teams chasing dashboards for six to eighteen months instead of fixing product-level content.
• The takeaway: use these tools for marketing, PR and crawl hygiene, but don’t confuse visibility with a growth strategy.

Summarises a Retailgentic piece by Scot Wingo in a good, bad and ugly structure, then offers a direct takeaway.

The critique targets tooling, not all AI search work; the same post credits crawlability improvements and sentiment monitoring as genuine uses.

For brands planning a campaign

A partner platform could focus on measuring how structured product content connects to commercial outcomes, rather than adding another AI visibility dashboard.

Examples: Source 1

Topics include ChatGPT instant checkout features, Reddit shopping, BigCommerce and Shopify earnings calls, and a predicted ChatGPT "universal shopping cart."

• The tone is enthusiastic, describing platform competition as a "battle of fiefdoms."
• One post is a two-part deep dive on Shopify’s agentic catalog with Shopify FAQs, attributed to Wingo and Retailgentic.

Short commentary posts summarising highlights and directing readers to the original through comment links, plus a two-part deep-dive format.

For brands planning a campaign

Clay can distribute news-driven analysis, but much of the framing originates with Retailgentic. A platform would need to agree on attribution and on what original analysis Clay adds.

Examples: Source 1 · Source 2 · Source 3 · Source 4

The Q&A post says agentic traffic already converts better than organic traffic and predicts it will surpass traditional search in under two years.

• Another post highlights a reported imminent ChatGPT "universal shopping cart," quoting Wingo’s view that many companies have tried a universal cross-merchant cart since 1999 without success.
• It predicts seven typical years of agentic commerce features will be packed into 2026, with an "epic Holiday 26."

Forecast posts drawing on Retailgentic reporting and quotes from Scot Wingo; the traffic-conversion and two-year claims appear directly in the Q&A post.

Treat the timelines as forecasts circulating in the field, not measured results.

For brands planning a campaign

Useful for a platform looking to frame the urgency around agentic commerce. The forecasts are attributed to a newsletter and a vendor, not independently evidenced.

Examples: Source 1 · Source 2

The post does not state whether they work for, own, advise or are otherwise paid by the company; ask them to clarify the connection.

View example

They credit Retailgentic as the source rather than claiming a role at the publication. Ask whether any sponsorship, client or ownership relationship exists.

View example

Both treat agentic commerce as already underway and lean on market numbers to make the case.

Van Camp writes from inside a vendor building the infrastructure.

• The posts announce Mirakl Agentic Activation and a joint research playbook with Merkle, and lead with headline figures: a scan finding under 1% of product pages above the readiness threshold, Salesforce's $67 billion Cyber Week influence figure, and 393% year-over-year growth in AI traffic to US retail sites.
• Clay curates others’ reporting, names sources, and is openly skeptical of AI visibility tooling as a growth strategy.

Read original · Original 2

Shivers also uses the teaching mode Clay adopts in the Q&A post: walk through what the standard does, then list what to do about it.

Shivers goes deeper on the protocols, contrasting OpenAI and Stripe's Agentic Commerce Protocol with Google's Universal Commerce Protocol and offering a feed-readiness checklist. Clay stays focused on product content and data hygiene, and is more likely to question the tooling market than explain a specification.

Read original · Original 2

Both link trust in the data to whether a listing is surfaced or skipped.

Joe frames the argument as a product pitch, proposing a 'Universal Compliance Protocol' and describing a compliance and trust engine.

• A second post tells a personal origin story about scanning product labels by hand at iHerb in the early 2000s.
• Clay comments and curates rather than building the infrastructure, and ends their Q&A post by referring readers to a service rather than a platform of their own.

Read original · Original 2

The two available LinkedIn posts are vendor-side thought leadership on agentic commerce.

• One announces Mirakl Agentic Activation, describing automatic enrichment of product content so AI platforms can read and recommend it, and direct connections between merchants and LLM platforms.
• It cites a scan of hundreds of product pages in which fewer than 1% cleared the threshold at which AI agents reliably surface products, and a Salesforce figure of $67 billion in Cyber Week 2025 sales influenced by AI agents.
• The other reports a joint Mirakl and Merkle research playbook, citing 393% year-over-year growth in AI traffic to U.S. retail sites and a projection that 55% of U.S. ecommerce spending will soon be agent-assisted.
• For a platform selling catalog or product-data infrastructure, Van Camp’s material already speaks in named launches and measured figures, giving a co-written piece a data-led narrative to build on.

The tradeoff is proximity to another vendor’s product story.

• Both posts advance Mirakl’s own capabilities, so a partner platform would co-write alongside a competitor’s messaging rather than an independent voice.
• The sharpest numbers are second-hand — a vendor scan and a Salesforce statistic — rather than the creator’s own measurements.
• The posts offer no skeptical counterweight: neither questions the category nor warns against chasing visibility dashboards.
• A platform seeking a candid “what does not work yet” angle would need to bring that perspective itself.
• Ask for audience figures, engagement data and commercial terms; none appear in either post.

Examples: Source 1 · Source 2

The two available posts compare competing agentic commerce standards.

• One explains OpenAI and Stripe’s Agentic Commerce Protocol as the earlier bridge that let ChatGPT consume product feeds, compare items, build carts, and check out using delegated payments while the retailer stays merchant of record.
• It then lists concrete adoption steps: provide a structured product feed to OpenAI, implement ACP checkout endpoints, support delegated payments, and keep pricing, inventory, and fulfillment logic consistent.
• The other explains Google’s Universal Commerce Protocol, framing the structured feed as the new homepage and the product graph as the new search results page, and closes by offering a feed-readiness checklist.
• A platform whose buyers are implementation and merchandising teams can use this instructional register directly, and the checklist offers a ready-made hook for a co-authored resource.

The tradeoff is a narrower register and thinner independent proof.

• Both posts explain protocols rather than report firsthand results, so a series with Shivers would rely on documentation and comparison instead of measured outcomes, and read as more technical and less brand-side.
• The material focuses on the OpenAI and Google rails rather than the messier internal reality of brand catalogs — split PIM, retailer, and direct-to-consumer feeds — where the question-and-answer format is strongest.
• Ask for audience data, engagement figures and documented commercial terms; the available posts provide none.

Examples: Source 1 · Source 2

One available post recounts starting an ecommerce career at iHerb in the early 2000s, scanning supplement labels and typing ingredients, dosages, and claims by hand, then running roughly 20,000 SKUs through the same pull-verify-update cycle.

• It draws a direct line from that manual accuracy to hiring and training from the warehouse floor.
• This is the only operational origin story among the alternatives, giving a platform a genuine narrative spine for a piece about why catalog work is unglamorous and load-bearing.
• The second post supplies a thesis to pair with it: after Google, Shopify, Amazon, Meta, Walmart, and Stripe aligned around agentic commerce protocols, the missing layer is a standardized compliance and trust layer, because agents do not troubleshoot and demand gets redirected to whoever has cleaner data.

The tradeoff is that the argument is attached to the creator’s own product.

• The compliance post pitches FeedZip.ai, so a partner platform selling data or trust tooling would be co-writing around a competing offer.
• The claim that trust is becoming the control layer of commerce is asserted more than demonstrated in the available text.
• The two posts also use different registers — a personal essay and an industry polemic — so a consistent multi-part series would take more shaping than one built on a repeating question-and-answer format.
• Ask for audience figures, engagement data and commercial terms; neither post provides them.

Examples: Source 1 · Source 2