When your customers ask AI,does your brand come up?

Does AI mention you at all?

We ask all three systems your category question from a clean account with no personalisation history. Asking from your own logged-in account is the most common way to get a falsely reassuring answer.

Measured: on the same question, A One Union was absent from all three clean accounts — but ranked second on a logged-in one.

What does AI say about you?

Presence is only half of it. AI reaches conclusions about your brand from whatever it has read, and your customers see those conclusions before you do.

Measured: ChatGPT described A One Union as “not a well-known major manufacturer” and said it “cannot be shown to be a top-tier brand.”

Who is being recommended ahead of you?

Ask the same question three times and you get three near-different lists. The order has not settled yet — which means whoever lays the groundwork first gets written into the answer.

Measured: three systems, five recommendations each, 13 distinct companies, only one appearing on more than one list.

AI brand check

What you get is a picture of where you stand:

• What all three systems say about you, verbatim
• Whether you appear in category questions, and where
• Every factual error, itemised
• The sources AI is reading you from

GEO growth programme

Fixing the reason AI cannot read you properly:

• Structured data and machine-readable layer on your site
• A brand knowledge base so every fact has an authoritative home
• Third-party content and media source placement
• Re-measurement each stage, before and after side by side

AI presence operations

Watching the gauge so you do not have to:

• Monthly measurement with a continuous record
• Competitor tracking — who is moving
• Continuous publishing and source maintenance
• Alerts when errors appear or positions shift

AI had already reached a negative verdict

Pressed further, ChatGPT called the company “not a well-known major manufacturer” and said it “cannot be shown to be a top-tier brand.” Customers see this. The company did not.

The name was wrong, and the wrong version was being cited

Its resort property was consistently misspelled in AI answers, because the systems were reading blog posts from 2022–2024 rather than anything official.

Personalisation lies to you

Asked from a logged-in account with relevant browsing history, the same question ranked the company second. That proves two things: supplying sources works, and self-testing is far too optimistic.

Why there is no “3× improvement” number here

Because this engagement is still at the diagnostic stage — the 90-day optimisation cycle has not run yet. When it has, and we have verifiable before-and-after measurements, they will go here. Until then this page only shows what we actually measured.

F Find (Measure)

Clean-account measurement across all three systems: where you stand and what is wrong.

L Locate (Trace)

Trace which pages, outlets and forums the answer was actually assembled from.

O Open (Build)

Put the correct facts into a shape AI can read, and place them where it already looks.

W Watch (Hold)

Re-measure, track competitors, alert on change. Positions move, so the station stays open.

Huang Chien-Tung — Founder

Co-founder of Chifeng Media. Years of media relations and brand communications work — the practical problem of what the outside world says about you. AI is the same problem with a harder counterparty.

YC — Technical co-founder

Formerly of Microsoft. Works in AI and large language models, distributed systems, data analysis and blockchain. Owns the measurement systems, data pipelines and audit trail.

About FlowEver

The Chinese name, 東流生生, comes from two ideas: a great river running east, and a line from the I Ching — “ceaseless generation is what change means.” A river that keeps flowing and keeps producing. Information behaves like water, and AI is its new outlet. You can ignore where it runs, or you can dig the channel.

What is GEO, and how is it different from SEO?

SEO puts you in a list of results and the user picks. GEO — generative engine optimisation — deals with the case where AI simply gives one answer, and that answer names only a handful of brands. If you are not named, you are not there. The techniques overlap, but GEO cares much more about factual accuracy and about which sources AI is reading you from.

Can you guarantee first place in ChatGPT?

No, and be careful with anyone who says they can. AI answers have no fixed ranking mechanism — the same question can return different results at different times or from different accounts. What we do guarantee: an auditable measurement method, before-and-after comparisons at every stage, and factual errors corrected.

How long before anything changes?

Factual corrections move fastest and can show up within weeks of establishing an authoritative source. Category recommendation lists move slowly; 90 days is a reasonable observation cycle. We re-measure at every stage rather than reporting only at the end.

Can you help if I do not have a website?

We can run the check, but optimisation will be limited. AI needs a source it trusts in order to treat something as fact, and your own site is usually the easiest one to control. Without one, the first step is normally to build that source.

Why can I not just ask AI myself?

You can, but use a clean account. Asking from your own logged-in account — the one that browses your own website — produces a personalised, over-optimistic answer. We hit exactly this in the A One Union case: second place from a logged-in account, completely absent from clean ones.