← Hub
DOC · YOREH-SCAN-01 REV 2026.07 · CONFIDENTIAL

Baseline Scan

AI Visibility
Baseline

What the AI models say about you today, before any of the work ships. This is the starting line we measure everything against.

Prepared by Node AI
Prepared for you, Fin Matson / yoreh.co
Issued July 2026
Scan ref 343643c5 · 20 July 2026
AT A GLANCE BASELINE READOUT
01
480
AI answers read
02
4
AI models
03
9%
Found unprompted
04
0 / 29
On Google's AI
The shape

Name Yoreh and the models know you. Describe the category and you are almost never there. And half the time they do describe you, they are describing an older version of the brand that your own site still shows them.


01 /

How We Read It

The method, in plain terms

We wrote 40 questions a real buyer might type, then asked each of the four big AI assistants every question three times, with live web access turned on. That is 480 answers. Half the questions never mention Yoreh by name (these test whether you get discovered when someone asks about the category). Half name you directly (these test how you get described once you are found). Separately, we checked Google's own AI answers (the AI Overview and AI Mode boxes) across 21 of your keywords.

The modelsChatGPT, Claude, Gemini, Perplexity
Questions40, across six product areas plus direct brand questions
Runs3 per question per model, so one-off answers do not skew the read
Answers read480 from the assistants, plus 29 Google AI results
Two lensesUnprompted (category questions) and named (brand questions)

We pinned the exact model versions so that when we run the same 40 questions again at week 6 and week 12, you are comparing like for like and the movement is real, not noise.

Read this honestlyTwo caveats

The competitor set is our pick, not a verdict. We chose Buffr, STRUGA and The Great Frog as sensible reference points. The models also surfaced others on their own, which we kept.

The Google AI read is the free tier. It captures whether an AI box appeared and who it cited, not the deeper paid signal. Treat the Google section as a snapshot, not the whole picture.


02 /

The Headline

Two problems, one bright spot
The Point

You have a recognition problem and an accuracy problem, not a reputation problem. When the models find you, they like you. The trouble is getting found, and getting the facts right.

96%
Name Yoreh, and the model describes you
9%
Ask the category question, and you appear
1 in 2
Named answers get a core fact wrong

Read those three numbers together and the whole scan is in them. The models have a rich picture of Yoreh, so nothing here is a cold start. But that picture only surfaces when your name is already in the question, and the picture itself is often out of date. The rest of this document is those two gaps, pulled apart so you can see exactly where they live.


03 /

Found When Named, Missing When Not

The gap, split by model

Every model shows the same pattern. Ask a brand question and you are described almost every time. Ask a category question, the kind a new buyer actually asks first, and you rarely come up. That gap between the two columns is the single clearest thing in the scan.

ModelNamed question, you appearCategory question, you appear
ChatGPT100%11%
Gemini96%11%
Perplexity96%8%
Claude92%5%
All models96%9%

The named column is strong across the board, which means the raw material about Yoreh is out there and the models can reach it. The category column is where the growth lives. Right now, when someone asks about smart ring jackets or silver jewelry without knowing you exist, they almost never leave the conversation having heard of you.


04 /

Where You Show Up, And Where You Don't

Unprompted discovery, by product area

This is the organic reach signal, and it is concentrated in exactly one place. When the question is about smart ring jackets, you surface about a third of the time. In every other area you make, you are effectively invisible to a buyer who does not already know your name.

Product area (category question)You appear 
Smart ring jackets (Oura, RingConn)35%
Recycled 925 sterling silver0%
Sourcing and sustainability0%
Alternative and unisex aesthetic0%

Across all the category questions that were not about smart rings, you were named exactly once in the entire scan. So the smart ring jacket is doing all of the discovery work on its own. That is a genuine foothold, the models already see you as a real answer there. It also means the recycled silver and alternative unisex stories are wide open, and someone else is currently getting the mention.


05 /

Which Yoreh The Models Think You Are

The name is doing three jobs at once

This is the part you asked us to watch most closely, and it is the clearest finding in the scan. Your name currently points at three different things, and only one of them is your jewelry brand. Which one a model lands on depends entirely on how the question is phrased.

When someone asks...Lands on jewelrySays BaliSays the Hebrew word
"What is Yoreh?" (bare name)9 / 120 / 1211 / 12
"...the jewelry brand Yoreh, where based?"12 / 1212 / 120 / 12

Read the top row first. Ask for your name cold, with no other context, and the models mostly answer with the Hebrew word (Yoreh, the early autumn rain, and a term from Jewish law), not your brand. So the default meaning of your own name is not you yet.

Now the bottom row, which is the bigger operational issue. The moment you tell a model it is a jewelry brand, every single answer places you in Bali, Indonesia, and none of them mention that you are a US registered company. Across all the brand questions, 61% put you in Bali, 19 answers describe a coffee shop or cafe attached to the brand, and not one identifies the US entity.

ChatGPT · brand questionThe brand is based in Bali, Indonesia, with a cafe and showroom attached to the brand.
Gemini · "what is Yoreh"Yoreh (Hebrew) refers to the first significant rainfall of the autumn season, and appears in Jewish law and rabbinic ordination.
Claude · brand questionThe jewelry brand Yoreh was founded by Matt Jacob and Fin Matson, photographers at heart... roles are somewhat divided.
Why this mattersAnd why it is fixable

The models are not inventing the Bali story. They are reading it off your own live pages. The manifesto and meet-the-team pages on yoreh.co still describe a Bali base and cafe, a 2022 founding and a co-founder. The models are quoting you accurately, they are just quoting the older version.

That is the good news buried in a bad-looking number. This is not a reputation you have to win back, it is a set of pages you control. Update what the site says, and you update what every model says next time it reads you.


06 /

When They Get It Wrong, What They Get Wrong

The recurring factual errors

Of the answers that name you, half repeat at least one fact that no longer matches the brand. The errors are not random, they cluster on the same handful of details, which is what makes them straightforward to correct at the source.

What the models sayWhat is trueHow often
Based in Bali / IndonesiaUS registered (Yoreh Supply LLC)61%
Founded 2022Founded 202324%
Co-founded by Matt Jacob and Fin MatsonFounded by Fin Matson18%
Run by an Indonesian entityYoreh Supply LLC (US registered)several
A broad gold jewellery lineGold only on the smart ring jacketsseveral

The pattern is consistent. Every one of these traces back to what yoreh.co currently tells a crawler, or to older press and directory listings the models found. Only three answers in the whole scan actually stopped to ask which Yoreh was meant. The rest committed confidently to the wrong version, which is worth knowing: the models do not hedge here, so the correction has to come from the source, not from hoping they get cautious.


07 /

How You're Described When They Know You

The bright spot

Set the accuracy issue aside for a moment, because the tone is genuinely good. When a model does describe Yoreh, it describes a brand it clearly rates. Out of every answer that mentioned you, one was negative. One.

80%
Of mentions are positive or better
59%
Of mentions actively recommend you
1
Negative mention in the whole scan

The words the models reach for are premium, distinctive, sculptural, solid, indie. They praise the design identity, the materials and the smart ring jackets specifically. This is why the headline calls it an accuracy problem and not a reputation problem. The sentiment is already an asset. The job is to make sure the model is being positive about the current, correct Yoreh, and to get it saying these things when your name is not in the question.


08 /

Who Wins The Rooms You're Not In

Share of voice, by pocket

In the category questions where you do not appear, the models still name someone. These are the brands getting the mention you are not. The list below counts how many times each came up, and in which area, across the scan.

BrandOwns the answer for...Mentions
The Great FrogAlternative and unisex silver, the reference point21
BuffrOura ring covers (also cited by Google's AI)12
STRUGAAlternative silver12
Also surfacedClocks + Colours, Chrome Hearts, Vitaly, Mejuri, Parts of Four, Emanuele Bicocchi1 to 2

Two things stand out. In the smart ring pocket, Buffr is not just winning inside the chat models, it is the brand Google's own AI names as best Oura cover. And in the alternative silver pocket, The Great Frog is the fixed reference the models compare everyone else against, including in answers that were literally asking how you compare. Those are the specific pockets where the category story needs to be told, because a name already fills the gap.


09 /

What The Models Are Quoting

The sources behind the answers

When the models cite a source, this is what they reach for. It tells you two things: where your current picture comes from, and which doors the category answers walk through without you.

SourceWhat it feedsCites
yoreh.co (your own site)Mostly returns, warranty, shipping and the manifesto page86
Oura supportSmart ring questions26
EtsySmart ring covers and jewelry21
RedditDiscovery and opinion, strongest on Perplexity~25
Competitor sitesThe Great Frog, Clocks + Colours, Mejuri, Parts of Four25+

The pattern is telling. Almost everything the models know about you comes from your own site, and mostly from your policy pages, not your story. Meanwhile the third-party sources that answer the category questions, the Reddit threads and the listicles, do not mention you at all. That is the exact target list for the off-site work: these are the pages that own the answers, and none of them carry your name yet.


10 /

Google's AI Surface

The AI Overview box

Across 21 of your keywords, you did not appear in a single one of Google's 29 AI results. Two things sit behind that number, and they point in different directions.

Reminder from the method: this is the free-tier read, so treat it as a directional snapshot of where the AI box exists and who it favours, not a full census.

11 /

What This Baseline Means

Where the work points, and how we'll measure it
The Point

Nothing here is a cold start. The models already know Yoreh and already like it. The work is to correct the record and widen the discovery, then re-run these exact questions to watch it move.

The scan lines up cleanly with the work already scoped in this hub, and it tells you which pieces carry the most weight.

We are deliberately not promising a specific ranking or a recognition percentage. What we are promising is a clear before, and an honest after. We run these same 40 questions, pinned to the same models, at week 6 and week 12. You will see exactly which numbers moved, in which pocket, and by how much. This page is the before.