Getting lighting products cited by ChatGPT, Perplexity, and AI Overviews

A specifier asks ChatGPT for a 4000K, 90+ CRI, DLC Premium troffer with 0-10V dimming, and your SKU never comes up - not because you don't stock it, but because your product page never told the model those four facts in a way it could use. That's the new failure mode in lighting distribution: not being out of stock, but being unreadable to the systems buyers now ask first.

Buying already moved to the chat box

Commercial lighting has always been a spec-driven category - CCT, CRI, lumens, wattage, beam angle, IP rating, DLC status - which makes it exactly the kind of product a specifier or contractor now types into an AI tool instead of a search bar. Gartner's own B2B research puts a number on the shift: a 2026 sales survey found 67% of B2B buyers now say they'd prefer a rep-free buying experience, continuing a run of Gartner findings that most of the B2B journey happens before a rep is ever contacted. Layer AI chat interfaces on top of that self-service instinct and you get buyers who ask an answer engine for a spec match, get three candidate SKUs back, and only then call a distributor to confirm price and lead time - if the distributor even made the list.

That's the part lighting distributors need to sit with. The AI didn't remove you from consideration. Your data did, before the buyer ever typed a word.

Why ERP-style lighting data disappears from AI answers

Most distributor catalogs still run on ERP-exported feeds built for order processing, not for being read by a language model. A typical row looks like LED TROFFER 2X4 40W 120-277V - a string a warehouse system can pick against, and a string an LLM has almost nothing to extract from. There's no CCT, no CRI, no delivered lumens, no DLC or Energy Star status, no dimming protocol. The part number is the whole product.

Answer engines aren't reasoning their way around that gap - they're pattern-matching on structured signal, and the data on this is fairly direct. An analysis of AI citation patterns found pages with structured data are cited roughly 3.1x more often in AI Overviews, and that 71% of pages ChatGPT cites carry some form of schema markup, while only a minority of e-commerce product pages implement it completely. Lighting distribution sits squarely in that gap: the specs a contractor cares about exist somewhere in a spec sheet PDF, but rarely as the discrete, labeled fields an AI system can lift into an answer.

Schema markup alone doesn't fix this. It's a delivery mechanism - it exposes fields, it doesn't manufacture values that were never captured. If your source data has no real CRI value, wrapping a blank field in JSON-LD gets you nothing. The fix starts upstream, with the values themselves.

What "ask an answer engine" actually looks like

Here's the kind of prompt a lighting buyer runs today - worth typing against your own catalog:

"Find a 2x4 LED troffer, 4000K, 90+ CRI, DLC Premium listed, with 0-10V dimming, from a distributor that has it in stock."

That sentence contains six filterable attributes. If your product data doesn't carry all six as discrete, correctly-valued fields, an answer engine has no basis to surface your SKU - it will cite whichever competitor's page states delivered lumens, CRI, and DLC status in a form it can parse.

Before and after: what enrichment changes

Raw ERP feed description, still powering most lighting distributor catalogs:

LED TROFFER 2X4 40W 120-277V 4000K

Enriched, quality-scored attribute table - what an answer engine can actually reason over:

The raw string carries wattage, voltage, and CCT if you squint. The enriched version carries every field a spec sheet or a contractor's RFQ would check - and each one is a candidate match point for an AI answer engine deciding whether to cite your SKU.

The DLC and Energy Star wrinkle

Lighting has a structured-data advantage most categories don't: DLC's Qualified Products List and Energy Star's certified list are themselves machine-readable, third-party-verified sources. A DLC ID or Energy Star certification can be extracted from those public lists, matched to your SKU, and pushed into your structured data - giving an answer engine two independent signals instead of one. Distributors who treat that status as a first-class, always-current attribute, not a PDF buried three clicks deep, are handing answer engines exactly the kind of verifiable fact they favor when choosing what to cite.

Where this gets fixed

None of this requires ripping out the ERP or the PIM, if there is one - plenty of lighting distributors run on spreadsheets and a website CMS, and that's a fine starting point. The gap is between what's stored and what's readable. Anglera sits on top of whatever you already run - Akeneo, Salsify, a flat file export, it doesn't matter - and does the work of gap-filling CCT, CRI, lumens, DLC status, and dimming compatibility from supplier documentation and spec sheets, scoring each value for confidence rather than guessing, then pushing the result out as clean, structured attributes. Most lighting catalogs can be running enriched within a few weeks, not through a multi-year systems overhaul, because the mechanism is enrichment on top of existing data, not replacement of it.

The category is getting quieter for the unprepared

This shift isn't lighting-specific, but lighting is an unusually good test case: the specs are standardized, and third-party verification sources already exist. The distributors who make their data machine-readable first will be the ones an AI Overview or a Perplexity answer actually names - and the ones who don't will simply stop coming up, without ever knowing why the quote requests slowed down.

That's the practical version of the AI-search shift for distribution: it's not about better marketing copy, it's about whether the facts a buyer needs are sitting in your data at all, in a form something other than a human can read. Anglera's job is making sure they are, continuously, at catalog scale, without asking a distributor to rebuild their systems to get there.