Home
ClikBy
ClikBy Knowledge Base
AI Selena

What Selena does not do

A short truth about product boundaries — so you do not promise the client extra and you set click protection correctly.

6–8 min Updated: July 2026
границыof the product
without mythsUI≠runtime
чеклиствместо

Key takeaways

  • AI Selena is an account for quality settings, a scoring profile, and (partly) retention; not a «magic antifraud button instead of the whole stack».
  • HVT/MVT/LVT thresholds, interests, and GA4/GTM fields in ai_selena do not by themselves cut clicks or send hits to analytics.
  • There is no product module «Golden Audience» and no guarantee of budget recovery from ads accounts.
  • Working detect = server bot + Smart Traffic SDK + visit statuses (+ optionally sync fraud in Ads).

Zone of responsibility

Honestly split three layers:

  1. Detect on the click — RedirectController::isBot, safe/soft_pass, Smart Traffic SDK.
  2. Selena account — UI thresholds, balance, interests, ideals score, sf_mode, site key, retention.
  3. Retarget — your pixels and Ads accounts; ClikBy sends a hit when fire conditions are met.

What is missing / it does not do

  • Does not classify every click by HVT/MVT/LVT thresholds in runtime (fields are saved, they are not read in the filtering pipeline).
  • Does not change isBot via Sensitivity Conservative/Balanced/Aggressive.
  • Does not filter traffic by selected Audience Interests from XML.
  • Does not automatically connect GA4/GTM/Yandex Metrica from Selena form fields alone (this is not a substitute for pixels on the link).
  • The links.ai checkbox («AI Selena traffic analysis») does not control redirect branches in the current code.
  • Does not create a «golden audience» entity in one click.
  • Learning Status from data.xml is a UI indicator, not a live model on every HTTP request.
  • Does not run dispute / chargeback in Meta/Google for you.
  • Is not a full replacement for enterprise ad verification (IVT/brand safety/CTV) — see the comparison on the help hub.
  • Does not promise a fixed «−X% CPA» from application logic — any % only from your measurements/cases.

A positive breakdown of detect capabilities: how they catch fraud.

What it is often confused with

ExpectationReality in the product
«Selena = all of antifraud»Antifraud on the click is broader: server + ST/SDK; Selena — settings and score
«I set Aggressive — it cuts harder»The field exists; a full Off/Aggressive matrix ≠ the full marketing UI
«Thresholds = autoban»UI thresholds ≠ autoban; look at visit statuses and sf_mode safe
«GA4 field in Selena = pixel»A google-analytics/GTM pixel is created in the pixels section and bound to the link
«Golden from interests»Interests are UI; the segment is assembled with a link+pixel+confirm

What to do instead of myths

  1. Turn on Smart Traffic on key links with pixels.
  2. Run safe → normal when you understand the would_block share.
  3. Watch Audience: bot_* vs confirmed.
  4. Collect «gold» as the practice offer→pixel.
  5. Read the limitations article together with the client report — fewer inflated promises.

What’s next?

Align expectations with the contour

Read fraud detect and the thresholds — and update the client brief.