Rankings alone do not describe every discovery surface
Generative engine optimization (GEO) concerns visibility in generated answers. It does not establish that search rankings are obsolete. Google’s guidance for AI features says foundational SEO practices still apply. Teams should investigate which surfaces matter to their buyers before changing targets.
Define what each metric measures
A mention records whether an answer names the brand. A citation records whether it references the brand’s domain or page. A session records a visit. A qualified opportunity records a business outcome. None is an interchangeable proxy for all the others.
The June 2026 Semrush AI Visibility Index release reports different citation patterns across four platforms in a US study covering January–April 2026. These are study-period observations. A platform average is not a performance target for every category.
Balance business outcomes and visibility diagnostics
What Matters recommends balancing leading and lagging indicators according to the team and objective. Business outcomes remain central at the top level. A mention share is an output of the measured AI systems, not an input the team controls. It may precede a commercial result, but calling it a leading indicator of that result requires a defined validation study.
My proposed operating approach is to separate the work, its visibility results and its business results. Improving source accuracy is work. A change in citation rate is an observed visibility result. A change in qualified demand needs its own measurement and attribution.
An illustrative OKR and measurement plan
Objective: Help qualified buyers evaluate our category and offering accurately.
- Business result: Agree a qualified-demand or conversion outcome with revenue leadership, using the team’s baseline and a defined period.
- Visibility diagnostic: Track mention and citation rates separately for a fixed panel of priority prompts, by platform. Set any improvement target only after observing baseline variation.
- Evidence-quality result: Reduce material errors in the approved public source set, with a documented review of each correction.
- Guardrails: Monitor organic visits, on-site conversion, measurement cost and content quality. Investigate trade-offs rather than assuming AI visibility replaces these outcomes.
Thirty prompts over an initial month is a suggested pilot scope, not a representative market sample or research standard. Record model, platform, account state, locale, dates, repeat runs and source URLs. Preserve outputs so a later comparison can be checked.
Assign ownership without claiming causality
The Semrush companion survey reported increased AI traffic or leads among 81% of organizations integrating SEO and AI visibility, compared with 36% managing them separately. This self-reported association does not establish that integration caused the difference. It supports investigating shared ownership, not promising the same result.
Assign research and analytics owners, agree how changes will be evaluated, and review progress at a scheduled checkpoint. Keep the original baseline visible when targets change. Do not reset targets simply to make progress appear stronger.
Key takeaways
- Keep business outcomes central.
- Measure rankings, traffic, mentions and citations as distinct signals.
- Separate source improvements from model outputs.
- Use a repeatable, dated monitoring protocol.
- Validate any claimed relationship with revenue before using it for forecasting.
FAQ
Should I drop rankings as a KPI entirely?
No. Retain rankings and traffic where they inform business decisions. Add AI visibility measurements when those surfaces matter to the audience.
How do you measure AI citations?
Record whether each sampled answer cites your domain or page. Track brand mentions separately, preserve the outputs and report results by platform, prompt set and date.
Is citation share a leading indicator of revenue?
Not automatically. It is a visibility outcome. Establish a relevant, validated relationship before treating it as a predictor of pipeline or revenue.
How often should these OKRs change?
Set review dates appropriate to the business cycle. Change targets when evidence or priorities justify it, document the reason and retain the original baseline.
FA