REP-320 · Lesson 1 of 12 · Module 1: How AI Forms an Opinion · 2:36

Where AI gets its opinion

An AI answer about a brand is a mirror of its footprint. Five source families feed it: your own site, reviews, news, knowledge bases such as encyclopedias, and forums and social posts. When they agree and are current the reflection is sharp; when they clash or go stale it warps. Every claim leaves a trail back to its sources, so trace a claim before you react. Different questions lean on different sources: feelings ("is it quiet?") lean on reviews and forums, facts ("what time is check-in?", "who owns it?") on your site, news and knowledge bases. The source mix tells you where to work. Worked example: two years after Quillmere Harbour’s renovation ended, assistants still warn about building noise, fed by 2023 reviews and a travel-forum thread, while the hotel’s own site never said the work was done.

Frame from lesson 1 of Sophyx Certified AI Reputation – Specialty, "Where AI gets its opinion"

What you will learn

  • Name the five source families
  • Trace a claim to its sources
  • Match each question to its sources

The one thing to remember

AI answers are mirrors. Change what they reflect.

Common mistake: “Our website is the source of truth for AI.”

To an assistant your site is one voice of five; reviews, news, knowledge bases and forums can outvote it, and did in the noise claim.

The fix: Your site is one voice of five: check what reviews, news, knowledge bases, and forums and social say, and work on the families that feed the claim.

Key terms

Source family
A kind of source AI draws on, such as reviews or news.
Stale source
A source whose facts were true once but are now out of date.
Source mix
The blend of source families behind one answer.

Transcript

The brands in the worked examples are fictional, and the numbers are illustrative.

Ask an AI about Quillmere Harbour, one of four boutique hotels, and it warns you about building noise.

That renovation ended in twenty twenty-four. Yet the answer still reflects it.

AI reputation specialty. Lesson one: where AI gets its opinion.

You will learn to name the five source families, trace a claim to its sources, and match each question to the sources it uses.

First, a quick recall from Answer Engine Specialist. Where does an AI answer come from?

From what the model learned in training, and from what it finds live, when you ask.

Think of each answer as a mirror. It reflects what the web says about you.

Five source families feed it. Your own site, reviews, news, knowledge bases like encyclopedias, and forums and social posts.

When they agree, the reflection is sharp. When they clash, or go stale, it warps.

Let's trace the noise claim. Which source most likely feeds it?

Reviews and a travel forum thread from twenty twenty-three. And the hotel's own site never said the work was done.

Every claim leaves a trail. Follow it to the source before you react.

Different questions pull from different sources.

Questions about how a place feels, like is it quiet, lean on reviews and forums. They set the tone.

Questions about facts, like check-in time or who owns it, lean on your site, news and knowledge bases.

So the source mix tells you where to work.

A common mistake: treating your website as the source of truth.

To AI, it's one voice of five. Check what the other four say.

Try it this week. Ask three assistants about your brand. List every claim, and tag each one with its source family.

Before the recap, can you name the five source families?

Your site. Reviews. News. Knowledge bases. Forums and social.

AI answers are mirrors. Change what they reflect.

Take the practice quiz. Next: facts and feelings.

Part of AI Reputation Specialty (REP-320)

For PR, comms and brand teams: find out what AI says about a brand, fix what is wrong at the source, and earn trust the honest way.

12 lessons, practice quizzes, a final exam and the Sophyx Certified AI Reputation – Specialty certificate. Regular price USD 89. Your first certification is free. Free with any paid Sophyx plan.

Start the course free
Book a Demo