Signals, not studies: what Optimizely’s failed FAQ test actually tells us about GEO

A sensible GEO tactic missed expectations. That is not a verdict on FAQs. It is a reminder that recommendations belong in a test plan—not a rulebook.

Build a Crib Sheet Read the CMSWire source

The signal

What happened

According to CMSWire’s September 10 report from Opticon 2026, Optimizely added FAQ content to pages because common GEO logic suggested question-and-answer formatting would help. The result underperformed expectations.

That is useful operational evidence. A plausible tactic was tried, and reality did not neatly follow the playbook. But the report does not disclose the pages tested, sample size, duration, comparison design, success metric, or numeric result.

The boundary

What it does and does not prove

It does signal

FAQ formatting is not a guaranteed GEO win. Treat it as a hypothesis and define the measurement before publishing the change.

It does not prove

That FAQs never work, that they caused a decline, or that they had a known effect on citations, traffic, conversions, or any other undisclosed outcome.

This is not presented as a formal peer-reviewed study. It is a company-reported test result, relayed by CMSWire, with important methodological details absent from the article.

Turn advice into evidence

The four-part GEO test card

Baseline

Record performance before the change so movement has a reference point.

Defined segment

Name the pages, query set, audience, or agent traffic being evaluated.

Measured outcome

Choose the observable result before testing—not after seeing the data.

Review window

Set the period and decision date so a tactic cannot run on belief forever.

Measurement checklist

Measure behavior, not belief

Citations and in-answer presence. Does the brand or page appear in generated answers for the defined query set?
Agent requests. What do server or CDN logs show about agent access and content retrieval?
Agent-mediated outcomes. Do agent-originated or agent-completed actions produce an observable business result?

One-company signal

The 6% figure needs a label

CMSWire reports that Optimizely’s own CDN logs showed agentic traffic to its website rising about 6% month over month during the interview period. That is one company’s operational evidence—not a market benchmark. The article does not define the absolute volume, exact window, or agent-classification method.