Classifying the Sources in an Answer and Finding Your Own Site

In this lesson

Lesson 1 introduced the source types. Now you'll apply them to real answers. First you'll classify one answer in five steps. Then you'll learn to read correctly whether your own site was cited. Finally you'll look at several answers together.

Five steps to classify an answer

1. List the visible sources. Take only the links the engine shows as sources. Note separately any web address written inside the answer text. It can be kept for reference, but it doesn't count as an engine citation. See citation evidence.

2. Label the type. Give each source one of the six types from Lesson 1: your own site, review platform, Reddit or another community, Wikipedia, news or trade press, comparison list. Use "other" for anything that doesn't fit. Don't invent new types on the fly.

3. Label the owner. Is the source yours, a competitor's or a third party's? A competitor's own site is its own case: it means the competitor's claims made it into the answer.

4. Match it to a sentence. Some engines show sources next to the sentence they support. Others list them only at the end. If you can match a source to a sentence, note it. If you can't, write "not matched" and don't guess.

5. Note the tone. Does the source say something positive, neutral or negative about you? This is your working note, not a metric.

Was your site cited? Three different cases

This takes more care than it seems. The three cases below are different:

• Cited. Citation evidence was captured and one source is your site.
• Evidence, but not your site. Evidence was captured but doesn't include your site. An answer with no sources from an engine that supports citation evidence also belongs here: it stays in the denominator as a miss.
• No citation evidence. The engine doesn't support citation evidence, or the evidence for this answer couldn't be captured. It stays out of the denominator and isn't counted as zero.

The third case doesn't mean "your site wasn't cited". There's nothing to measure. Definition: Owned-Site Citation Coverage.

A reminder: being cited doesn't mean being recommended, and it may not even mean being mentioned. Module 1, Lesson 3 separated those three ideas.

Worked example (representative)

The answers below are fictional. "Acme Helpdesk" is a fictional product, the questions were written for this lesson, and the numbers come from no real measurement.

Here are the sources behind five answers that mention Acme Helpdesk, all from the same week. A, B, D and E came from the same engine. C came from another engine that doesn't expose citation evidence.

• Answer A: a G2 category page, a Reddit thread, acmehelpdesk.example/pricing
• Answer B: a "best help desk tools" comparison list, a Reddit thread
• Answer C: no citation evidence (this engine doesn't expose it)
• Answer D: a trade-press article, acmehelpdesk.example/customers
• Answer E: a G2 reviews page, a Wikipedia article on help desk software, a competitor's own site

Type breakdown (10 sources across the four answers with citation evidence):

• Review platform (G2): 2
• Reddit: 2
• Own site (Acme Helpdesk): 2
• Comparison list: 1
• Trade press: 1
• Wikipedia: 1
• Competitor's site: 1

Owned-site citation: Answers A, B, D and E have citation evidence, so that's 4 answers. Two of them (A and D) cite Acme Helpdesk's site. Owned-site citation coverage is 2 ÷ 4 = 50%.

Answer C stays out of the denominator. If you counted it as "not cited", you'd get 2 ÷ 5 = 40%. That would be wrong, because C had no evidence to measure.

Had C come from this same engine with an empty source list, it would stay in the denominator as a miss, and 2 ÷ 5 = 40% would be correct.

What the breakdown tells you, and what it doesn't

What you can take from it:

• In this sample, the answers that mention Acme Helpdesk lean on review platforms and Reddit as much as on its own site.
• Only the pricing and customers pages were cited from Acme's own site. The product and documentation pages never appear, which is worth looking into.
• Reddit shows up twice. Check which sentences those threads support.

What you can't take from it:

• Five answers from one week is a small sample, not the overall picture. As Module 2 showed, a single sample can mislead.
• It doesn't explain why a source was picked.
• It doesn't show pages the engine read but didn't display.
• Being cited doesn't mean being recommended.

Practice: keep a source log

A simple table is enough to classify sources the same way every time. Suggested columns: date, engine, question, source URL, type, owner, supported sentence, tone.

Two rules keep your logs comparable:

• Use the same question set each time.
• Measure under similar conditions: the same engines and a similar time window.

If you change the question set, old and new breakdowns can't be compared directly. Module 6 goes into this.

Summary

• Classify an answer in five steps: visible sources, type, owner, supported sentence, tone.
• A web address written inside the answer text isn't an engine citation.
• "Not cited" and "no citation evidence" are different. An answer with no sources from an engine that supports citation evidence stays in the denominator as a miss. Answers whose evidence couldn't be captured, and answers from engines without citation evidence, stay out of the denominator and aren't counted as zero.
• In the representative example, owned-site citation coverage is 2 ÷ 4 = 50%. Counting C, from an engine without citation evidence, as zero would wrongly lower it to 40%. Had C come from this engine with an empty source list, 40% would be correct.
• A source breakdown gives you clues. It doesn't prove causes, reveal the hidden pool or generalize from a small sample.