Run an AI visibility audit for a client

What should the audit agreement contain?

Define the brand, website, relevant product versions, target markets and the decisions the client wants to improve. Ask for approved documentation and the owner who can resolve disputed claims. Include a short list of actual alternatives, but keep the comparison fair to the buyer’s constraints. A client that serves one country should not be assessed as if it offered every market.

Agree what will be delivered and what requires additional access. The audit might include a question inventory, reviewed answers, a citation ledger and a prioritized correction plan. Analytics or CRM evidence needs its own access and scope. Set the review date and number of scenarios before collection so the final sample cannot be quietly narrowed to the most dramatic answers.

How do you establish the baseline?

Use questions from interviews, sales objections and procurement documents. Separate category discovery, named comparisons and factual checks. Record engine, interface, language, region where supplied, search mode and date for every run. Retain failed requests and excluded items with reasons. Preserve complete answers so reviewers can inspect the context and unfavorable findings.

In Rankfor, open the authorized client project and inspect Rankfor Index for completed scans. Use a separately launched Answer Trail for a custom question. Investigate material variation with Dice Roller Test where the account has access. The public playbooks show role-based examples; your evidence must identify the actual records used for this client.

How do you turn observations into findings?

Check factual claims against the client’s approved evidence and inspect whether cited pages support the associated statements. The citation-verifiability study provides the distinction between citation presence and support. Label errors, omissions, unsupported citations and legitimate unfavorable comparisons separately. A competitor recommendation can be reasonable when it better satisfies the prompt.

Consider an illustrative audit for the fictional Northstar service. One answer uses an obsolete cancellation fee and cites an old directory. The current policy gives a different rule. The finding contains the exact claim, both URLs, relevant dates and the affected buying decision. The action is to verify and correct owned material and request a substantiated listing update.

The evidence does not prove that the directory caused the answer or that removing it would change the model.

Which finding should guide the next cycle?

Chapter 8 of Become the Answer groups findings into five diagnostic rows. Use presence in buyer questions, factual and message accuracy, useful page coverage, position in cited external sources, and consistency across markets or dates.

For each row, record the observation, comparator, rule for concern and next check. Select one primary row for the next cycle. For example, accurate owned pages alongside recurring external sources that omit the brand justify investigating those sources. That pattern locates a weakness; follow-up checks still need to test why it occurs before the team selects a remedy.

What belongs in the handoff?

Give each action an owner, an effort estimate and an acceptance condition. Choose among updating a page, consolidating conflicting material, producing missing evidence, creating a new asset or taking no action. Review Brand DNA Map and Personas > Cluster Coverage as inputs, while labeling generated audiences and gaps as hypotheses.

Open account-level Reports, select the correct project and inspect the report before exporting PDF, Markdown or CSV. Attach your evidence ledger and prioritized plan as separate audit materials when needed. Do not promise a particular report includes every custom annotation automatically. Finish with a review meeting that resolves disputed facts and approves the next work, while keeping expected outcomes conditional on later evidence.