What AI Actually Recommends When You Ask for Non-Toxic Cleaners

Why we ran this

More and more parents don't start a cleaning-product search on Google anymore — they ask an AI assistant "what's the safest cleaner for a house with a baby?" and take the answer at face value. So we wanted to know, honestly and with receipts: when you ask AI that question, whose sources does it read, and whose brands does it name?

We're a cleaning brand, so we'll be upfront: we have a horse in this race. That's exactly why we're publishing the raw method and every number, including the parts that don't flatter us. The goal is a useful public dataset, not a trophy.

How we did it (the method, in plain English)

• We wrote 30 plain-language questions a real parent would ask — for example, "best non-toxic all-purpose cleaner," "safest cleaner for homes with babies," "best pet-safe non-toxic floor cleaner," and per-room and per-concern variants (pregnancy, asthma, fragrance-free).
• We sent each one to gemini-2.5-flash with live web-search grounding turned on (Google Search grounding), the same class of system that powers AI answers people now rely on.
• For every answer we logged two things: the source domains the AI cited and the cleaning brands it named (matched against a fixed list of 24+ brands).
• 30 of 30 queries returned a grounded answer. Every raw response is saved on file.

Honest caveats: this is one model on one day, not every AI. Brand detection is keyword-based, so a brand mentioned only to be dismissed still counts as a "mention," and a brand named in an unusual way could be missed. Grounded search results shift over time. Treat these as directional signals of AI visibility, not a safety ranking.

Finding 1: AI reads a surprisingly small set of sources

Across 30 answers, the AI cited 161 distinct domains — but a handful did most of the work. Here are the most-cited sources.

The pattern: environmental databases and independent review publishers dominate, with community forums (like Reddit threads) showing up repeatedly. Brand-owned pages rarely win the citation. If you want AI to recommend you, being written about by these sources matters more than what's on your own homepage.

Finding 2: AI names a familiar short list of brands

There was no single "AI's favorite cleaner." Instead, the same cluster of legacy clean-cleaning brands surfaced again and again. Here's how often each brand appeared in an answer.

The takeaway for shoppers: an AI naming a brand a lot tells you it's visible in the sources above — not that it's been independently verified as the safest. Two brands can look equally "AI-approved" while having very different ingredient lists. That's exactly why we built a plain-English ingredients glossary: so you can check the label yourself instead of outsourcing the decision.

What this means if you're the one asking

Three practical lessons from the data:

1. AI is a librarian, not a lab. It repeats what a few trusted sources already say. If those sources are thin on a brand, the brand vanishes from the answer — good or bad.
2. "Recommended by AI" and "verified safe" are different claims. Ask the follow-up: can I read every ingredient? A brand can be popular in AI answers and still hide its full formula behind the word "fragrance."
3. The label still wins. The most reliable move isn't asking a machine which bottle to trust — it's learning to read the back of the bottle. Here's what "non-toxic" actually means on a label.

Where Ecolosophy fits (the honest part)

We're newer than most of the brands on that list, so we don't expect to top an AI-visibility chart yet — and the data reflects that. What we can control is being the kind of source worth citing: publishing every ingredient, showing our work, and telling you when a claim is a myth (SLS isn't a carcinogen) versus a real concern (formaldehyde-releasing preservatives are). One plant-based concentrate makes 100+ spray bottles, every ingredient is named, and there are no artificial scents or synthetic chemicals. Family-safe, pet-safe, planet-safe.