Explaining LLM
An LLM, or large language model, is a neural network trained on vast text to predict the next token, the engine behind ChatGPT, Gemini, Claude and AI search.
Decoded
Reference articles that each go deep on a single ad tech term. Every entry in the Explaining series opens with a plain-language definition, then covers the mechanics, origins, key dates, limitations, and commonly confused adjacent terms - grounded in PPC Land coverage, platform documentation, and industry standards from bodies like the IAB.
Posts
Previous Posts-
-
Explaining AI referral traffic
AI referral traffic is website visits that arrive when a user clicks a link inside a chatbot answer, such as ChatGPT, Gemini or Perplexity. Small, fast-growing.
-
Explaining collective redress
Collective redress is the legal route letting one representative sue for many people harmed by the same conduct, as in the EU Representative Actions Directive.
-
Explaining qualified entity
Qualified entity: an organisation designated by an EU member state to bring representative actions for injunctions or redress on behalf of groups of consumers.
-
Explaining pay or okay
Pay or okay: a consent model in which users either accept tracking for personalised ads or pay a fee to avoid it, contested in the EU under GDPR and the DMA.
-
Explaining standard contractual clauses
Standard contractual clauses are EU-approved contract terms that let firms send personal data to countries lacking an adequacy ruling under GDPR Article 46.
-
Explaining right of access
Right of access is a person's legal right to learn whether a company processes their personal data and get a copy, set out in GDPR Article 15 and US state laws.
-
Explaining sellers.json
Sellers.json is the IAB Tech Lab file in which ad exchanges and SSPs list the publishers and intermediaries they pay, so buyers can identify the real seller.
-
Explaining app-ads.txt
app-ads.txt is the IAB Tech Lab file in which app developers list the ad systems authorised to sell their inventory. How it works, where it fails, what changed.
-
Explaining debtor-in-possession
A debtor-in-possession is a company in Chapter 11 that keeps running its own business with a trustee's powers, under court supervision, while it reorganises.
-
Explaining stalking horse bid
A stalking horse bid is the first, pre-negotiated offer for a bankrupt company's assets, setting the floor price rivals must beat at a court-supervised auction.
-
Explaining observability
Observability is the ability to infer a system's internal state from its outputs - logs, metrics and traces - used to debug ad auctions, tags and AI agents.
