Trust & standards

AI can help organize drafts. People decide.

WhiskerMatch uses AI carefully. AI may help draft and organize information, but people review profiles, applications, and placement decisions. Trust is not just a policy page. It is the product behavior: hold, review, source, approve, and record why.

Check if this fits your team See how human review stays in control

Ten behaviors a team can inspect.

The product should make its boundaries easier to see than the policy language describing them.

01Human decisions
02AI boundaries
03Review gates
04Source trail
05Unknown fields
06Public profile safety
07Applicant fairness
08Commerce separation
09Data and privacy
10Proof permission

AI can help

draft, structure, summarize, and flag missing fields

People make the decisions

staff review profiles, applications, and placement calls

Never automated

no auto-placement, no auto-rejection, no pay-to-rank

Missing stays missing

unknown fields are not filled with guesses

Named review required

Nothing public without review. A real person is attached before publishing

What WhiskerMatch will never do

Trust starts with the lines that do not move.

Early teams should not have to infer the boundary from a policy page or trust a product demo on faith. These are operating constraints, stated plainly enough to inspect.

We do not auto-approve adopters.

An application can be organized for review. Approval remains a shelter or rescue decision.

We do not auto-reject people.

No model output closes an application or sends a rejection on its own.

We do not hide AI reasoning from reviewers.

Prepared work stays labeled, its source stays visible, and the reviewer can see what needs confirmation.

We do not pretend every animal is easy to place.

Unknowns, edge cases, and changing context are part of the record—not friction to erase from the story.

We do not turn adoption into a score alone.

A score cannot carry policy, lived observation, exceptions, relationships, or professional judgment.

We do not ask early teams to trust a black box.

The pilot begins with inspectable sources, named reviewers, visible holds, and a record of the final action.

We do not invent proof before real pilots exist.

No borrowed logos, implied customers, fabricated metrics, or testimonials written in advance.

Anti-black-box boundary

Special-needs animals never become a machine score.

Photos, care notes, medical sources, foster observations, and behavior context may help a reviewer understand the record. None of them gives AI authority over an adopter or placement.

AI does not approve adopters.
AI does not reject adopters.
AI does not make final placement decisions.
AI does not diagnose animals.
Photos are context, not decisions.
Medical and behavior notes require human interpretation.
Reviewers can override or stop prepared work.
Edge cases belong to accountable people.
Special-needs animals are never reduced to a score.
Uncertainty stays visible until a person resolves it.

AI governance

What AI does. What humans do. What we never automate.

AI drafts, structures, and summarizes. It does not make placement-level decisions. These lines are product constraints, not preferences.

SurfaceAI assistsHumans ownNever automated

Animal profiles
AI
Drafts structured fields from intake sources as a review aid
Human
Reviews, edits, and approves before publication
Never
Never invents fit claims or medical status; never scrapes or pads a profile to look more complete than the record is
Workflow state
AI
Surfaces reviewed status, freshness, and shelter-defined workflow state
Human
Owns placement-level decisions
Never
Never produces an opaque applicant score or manufactures certainty the shelter does not have
Care & behavior content
AI
Drafts from a curated source library for human review
Human
Reviewed by credentialed reviewers with dates and sources
Never
Never replaces a veterinary relationship and never gives medical, legal, or compliance advice
Photos & special-needs context
AI
Keeps photos attached to source notes and prepares care context for review
Human
Interprets daily needs, uncertainty, and every placement question
Never
Never diagnoses, decides from a photo, or reduces a disabled animal to a risk score
Applications
AI
Groups and summarizes for human review
Human
Triages, responds, and decides
Never
Never auto-rejects and never auto-places based on model output
Review queues
AI
Organizes by reviewed status, freshness, and shelter-defined workflow state
Human
Sets workflow rules and decides every priority
Never
Never accepts pay-to-rank and never ranks adopters for placement

Essentials and partner offers are post-placement only — isolated from the placement surface

Prevents

Commerce nudging an adopter toward an animal or a shelter

See these boundaries in motion in the interactive product demo — each event shows what AI may draft and what a human must review.

No-pay-to-rank architecture

Commerce and placement are separate systems, by construction.

This is not a promise to behave. There is no path for money to reach ranking, discovery, matching, placement visibility, or applicant priority.

Placement system

Animal record
Placement workflow
Public profile
Applicant workflow
Discovery / ranking

Commerce system

Essentials catalog
Sponsors
Partner resources
Post-placement support

Commerce cannot influence animal ranking · discovery order · matching · placement visibility · applicant priority · reviewer priority. Different systems, different incentives, no shared path.

When information changes

A record is not frozen the moment it is published.

Animals change. Intake notes get corrected. Behavior is re-observed. The point of one reviewed animal record is that change is tracked and visible — not silently overwritten.

New information arrives

A staff member, foster, reviewer, or adopter flags something — a medical update, a corrected age, a behavior change. The flag routes back into the queue with an owner, not into a comment that nobody reads.

The record is re-reviewed

A human reviews the change before it changes the public record. Medical and behavior content has a standing review cycle. Reviewed is dated, so a stale review is visible as stale.

The change is attributed

What changed, who changed it, and when is recorded on the record. A correction is a tracked event, not an invisible edit. The history travels with the animal.

Uncertainty stays honest

If something is no longer known with confidence, the field returns to unknown rather than holding a value that is no longer true. Unknown stays unknown — including after a change.

Data deletion & export posture

What we hold, what we don't, and how you get it back.

Stated plainly so it can be checked. This describes the posture, not a substitute for the written terms of a pilot agreement.

Pilot requests are received and stored, not just drafted

Submitting the pilot form at /pilot sends your answers directly to WhiskerMatch's database — it's not just an email draft prepared in your browser, though a direct mailto: to cazar-brandes@outlook.com remains available if you'd rather use that.

Deletion requests

A deletion request for a pilot request or a provisioned organization's data can be made by email to cazar-brandes@outlook.com and is handled by a person.

Product data export

For organizations in the pilot, structured export of your records is part of the pilot agreement. If you decide not to continue, your data leaves with you. No lock-in is the intended posture.

No quiet secondary use

Organization and household data is not sold, rented, or fed into a marketing or ad graph, and is not used to train third-party models. It stays inside the system it was given to.

This is the stated posture for an early, founder-led stage. The binding specifics — retention, export format, deletion timelines — live in the pilot agreement, not in marketing copy.

What “reviewed” means

Review is a process, not a stamp.

On WhiskerMatch, “reviewed” has a specific operational meaning. The parts travel with the content so a reader can see them without clicking through to a policy page.

Who reviews

Credentialed humans with a domain match — DVMs on medical, certified trainers on behavior, shelter staff on animal-level fit. Credentials are attached, not implied.

What gets attached

Reviewer name and credentials, the date of last review, the sources consulted, and the region the review applies to. All visible on the content.

What triggers re-review

Flags from staff, reviewers, or adopters route back to queue. Medical and behavior content has a standing review cycle — reviewed isn't set-and-forget.

What it isn't

Not a stamp. Not AI output relabeled as human work. Not a one-time approval that persists forever. AI drafts are review aids; humans decide what ships. Review is dated for a reason.

Credentialed reviewers are vetted by the founder, not self-claimed. If you hold relevant credentials, the reviewer path is open.

Apply as a reviewer →

Trusted expert notes

Read access broadly. Write access selectively.

Expert notes are a high-signal annotation layer on care, behavior, and medical context — not an open forum. Most users read. Only vetted contributors write.

Who can write

Credentialed DVMs, certified trainers, shelter-medicine specialists, and other vetted welfare experts. Credentials are verified by the founder, not self-claimed.

Who can read

Anyone viewing an animal profile or care article. Expert notes are a public layer — the barrier is on publication, not consumption.

What gets published

Care guidance, behavior context, medical background, and placement-relevant notes — all reviewed, dated, and attributed. Not opinion threads. Not anonymous tips.

What it is not

Not a social network. Not an open comment section. Not animal welfare Reddit. Not a place for anyone with an opinion to post. High signal, low noise, by design.

Future reviewer standard

There is no live reviewer network to claim today. Any future medical, behavior, or care education would require relevant credentials, identity checks, domain matching, clear attribution, and accountable human review before publication.

Review states live on the content

You never have to guess what you're reading.

Care, behavior, and medical context carry state, reviewer, last-reviewed date, and sources on the page itself — not buried in a policy footer somewhere.

Reviewer credentials attached to every reviewed piece
Last-reviewed date and sources shown inline
Flagging is frictionless, routed to re-review
Region-aware, not globally flattened

Care · Dogs · First week home

Reviewed

Decompression in the first 72 hours: what actually helps.

A short, evidence-aligned guide for the first three days home — sleep, space, food, and what to watch for. Written to support a veterinary relationship, not replace one.

Reviewer

Dr. A. Ngata, DVM

Last reviewed

Mar 18, 2026

Sources

4 linked

Illustrative. State, reviewer, and sources travel with the content.

Lines that do not move

Platform-wide standards that aren't up for negotiation.

These apply to every surface of WhiskerMatch. They are enforced defaults, stated in public so behavior can be checked against them.

No pay-to-rank, ever.

No partner, brand, shelter, or advertiser can pay to change placement order. Not on animals, not on essentials, not on anything that touches discovery.

No fabricated animal facts.

If a detail isn't confirmed by the shelter's intake or a named reviewer, the field stays unknown. Unknown is a valid state.

AI drafts are review aids. Humans decide.

Drafting, summarizing, and normalization are automated. Every placement-level decision — adopt, foster, publish, reject — stays with a named human who knows the animal. AI output is not a substitute for human judgment.

The shelter owns its record.

Organizations keep authority over how their animals are represented, which fields are public, and where their data goes. The platform doesn't override that authority.

Educational content carries its sources.

Care, behavior, and medical context show the reviewer, the review date, the sources consulted, and the region the guidance applies to — inline on the content, not buried in a footer.

Commerce stays separate from placement.

Curated essentials and partner products never influence which animals appear first, which profiles get prominence, or which shelters are recommended. Different systems, separate databases, no cross-contamination.

Expert notes are read-broadly, write-selectively.

Care and behavior guidance is written by credentialed, vetted specialists — not crowdsourced, not open-posting. Read access is public. Publication is controlled. High signal, low noise.

No dark patterns.

No countdown pressure. No manufactured scarcity. No guilt copy. No default-opt-in tricks. The product treats staff and adopters as adults.

What WhiskerMatch is not

Three hard boundaries.

The principles above describe what the platform enforces. These three describe what it refuses to become — no matter who asks.

Not a veterinary authority.
WhiskerMatch is not a veterinary authority. The platform does not diagnose, prescribe, or replace the veterinary relationship between an adopter and their clinic. Care and medical content is drafted from a reviewed source library and carries the reviewer's name, credentials, and review date. It is provided for informational purposes only and is not medical advice. It is not a substitute for a professional clinical examination, diagnosis, or treatment. If an animal is experiencing a medical emergency, contact a veterinarian or emergency clinic immediately.
Not a ranking auction.
No partner, brand, shelter, or advertiser can pay to change placement order, anywhere on the platform. This is a platform-wide constraint, not a line item in a pricing tier.
Not a silent data extractor.
Household profiles and shelter records are not sold, rented, or quietly fed into a marketing or ad graph. What an adopter shares with a shelter stays between that adopter and that shelter.
Not an autonomous decision-maker.
WhiskerMatch does not use AI to make placement decisions, auto-reject applicants, or publish records without human review. AI drafts are review aids. Humans retain full authority and accountability for every decision that affects an animal's placement.

Founder-supported pilot

Bring one workflow and inspect the boundaries together.

Start with one workflow, named reviewers, attached sources, and no public change without human approval.

Talk through a pilot See the review flow