IN-HOUSE PRODUCT / FWDLink
FwdLink. See how AI answers describe your brand.
FwdLink is an AI visibility monitoring and improvement service that HYEROK IT builds and runs. It asks AI the buying questions a brand selects and shows, with the recorded answers, whether the brand was mentioned, which pages were cited and which competitors appeared.
The problem we saw
Buyers now ask AI which product to choose, not only a search box. But a company rarely knows whether its brand appears in those answers, or what a competitor was recommended on.
So instead of a vague visibility score, FwdLink asks the buying questions a team chose under recorded conditions and keeps the raw answers and the pages they cited.
Being left out of an answer is not treated as proof of a cause. The answer is set beside the facts the team confirmed and its public pages, and a person decides what to change.
What we built
How one question moves from measurement to improvement and back to measurement.
-
Choose questions, collect answers
Once the brand, domain, market, language and buying questions are confirmed, the same questions go to the OpenAI and Google Gemini search APIs several times. Each answer keeps its collection status.
-
Mentions, citations, competitors
Per question and engine, it shows whether the brand was mentioned, which pages were actually cited and which competitors appeared. Every number opens onto the answers behind it.
-
Drafts from confirmed facts
Answers sit beside the product facts the team confirmed and the live public page, with a draft written only from confirmed facts to compare. Drafts are copied or exported; FwdLink does not publish them.
-
Record the change, measure again
The date, URL and content of a change are logged, then the same questions, engines, repeats, market, language and search settings are observed again. If a condition differs, it says so instead of showing a change rate.
What mattered in the design
We focused on showing AI answers without overstating them, in a way anyone can check against the evidence. Expand one to read the reasoning.
-
A mention is not a recommendation
Being named in an answer is different from being recommended. A person judges recommendations from the exact sentence, and a failed answer is never counted as "not mentioned".
Numbers alone are easy to misread, so every number opens onto the answers behind it.
-
API observations are not user impressions
FwdLink collects answers from the OpenAI and Google Gemini search APIs. They are not a count of consumer-app users or impressions, and any consumer-app checks are labeled separately.
A result can only be trusted when it is clear what was measured.
-
Observation and hypothesis stay apart
The observed answer, the page check and the improvement candidate go in separate fields. A missing mention alone is never presented as the reason.
Writing a hypothesis as a cause leads teams to fix the wrong thing.
-
Drafts use confirmed facts only
When a needed fact is not confirmed, FwdLink asks the team for it instead of writing it. The client reviews and applies a draft themselves; FwdLink does not change public pages.
AI writes plausible sentences easily, so something had to stop unsupported claims from reaching a public page.
-
Compare only under the same conditions
Before and after are compared only when questions, engines, models, market, language and search settings match. Otherwise it says so instead of showing a change rate, and a difference is never claimed as caused by the change.
AI answers vary from one ask to the next, so a comparison without matched conditions supports no decision.
How it runs today
We build, deploy and operate the product, and handle customer support ourselves.
- Available now
- AI visibility checks, plus scheduled monitoring that observes the same questions again every week. After creating an account and confirming the brand and questions, the customer starts a check, and results appear answer by answer in a private workspace.
- Early access
- The improvement workflow: comparing answers with confirmed facts and public pages, reviewing drafts, logging changes and observing again. Agency and team workspaces for several clients are at the same stage.
- What a check needs
- Brand name and domain, a one-line description, language and market, and buying questions. Competitors are optional, and no other system needs to be connected.
- What it does not promise
- It does not guarantee appearing in or being recommended by AI answers. It helps teams review and update information, then observe what changes under comparable conditions.
- Support
- HYEROK IT handles enquiries, account support and improvement work conversations directly.
Need something similar?
If you are planning a tool that collects AI answers and shows the evidence, a workflow where people review and decide, or a service you run yourself, tell us which features you need.