What results should an AEO contract measure?

How should you count your brand in AI answers?

I would judge an AEO contract first by whether it makes each result clear enough to check. Start with whether an answer names your brand for a question a real buyer might ask. A mention is a name in the answer. A citation is a link to your page or a named source. They are related, but they are not the same result. A page can be linked without the answer naming your brand. An answer can name your brand without a link to your site.

Write both definitions into the contract. State which AI products are in scope and which parts of an answer count. Say whether a name in the answer text counts, whether a name only in a linked source counts, and how you will handle an answer that does both. Keep the counts apart so a reader can see what happened.

For a worked example of a brand name and a source link appearing in the same answer, see the guide on what Gemini checks before citing a business. The two observations belong in separate columns, even when they appear together.

Save the exact question, answer, product and date. Record the place and account settings when they matter. Keep each later answer as a separate observation, with its own date.

Agree on a baseline before work starts. A baseline is the starting count taken under the same rules you plan to use later. If the question set or test conditions change, show the old and new sets apart. Record the baseline date and name who can review a disputed entry. Review the baseline and later observations at the monthly meeting.

Which buyer questions should the contract cover?

The question list gives the count its meaning. A brand mention on a question far from your work may be less useful than a mention when a buyer asks for help you sell. Ask for a list that reflects the services, places and buyer needs in the agreement. It should include both broad questions and questions close to a buying decision.

Group questions by service and market, then separate provider-choice questions from broader questions. This lets your team see which buyer needs have a qualifying mention.

For each question, record the exact words and the group it belongs to. Agree on the products to check, the places represented and the dates for review. If your business has several brands, specify which name belongs to each group. A mention of a parent brand should not silently count as a mention of every local brand.

You can agree on a fixed core list and a separate list that changes as the market or your offer changes. The fixed list helps you compare like with like. The second list helps you see new questions. Mark additions and removals so a new question does not look like a gain on an old one. Do not let either side swap out hard questions without a written note.

Report the number of questions with a qualifying mention beside the total checked. Include the questions with no observed mention. Review the fixed list monthly and agree on its size before the baseline. The fixed list gets one review each month.

How should you count leads from AI search?

A lead is a person or firm that has taken an agreed step toward buying. The contract should say which steps count. A form sent, a booked call and a phone inquiry can each mean something different. Choose the events your team can check and define when sales will treat a lead as qualified.

Count a contact form under the agreed lead rule. When sales checks whether the buyer needs a service you offer, record the qualified status separately. Keep both stages in the report.

Name the system that holds the count and who owns it. Set rules for repeat forms, existing customers, spam, staff tests and leads outside your service area. Agree on the date used for each stage and how late sales updates will be added. A lead should not count again merely because its status changed.

Google Analytics calls actions that matter to your business key events. Google says that Google Ads conversions and Analytics key events have distinct reporting rules. If you compare sources, use the same lead definition and the same system for the comparison. Otherwise, a difference in the report may come from the way events are counted.

Source is separate from lead quality. Record a tagged visit when available, and keep a buyer's answer about how they found you as a reported source. Record the question your team asked.

Keep the source evidence beside each lead and distinguish a referral from a buyer’s own report. The guide to AI search cost per lead explains the available referral and search reporting in more detail. Review qualified leads once a month.

How should results be reviewed over time?

The contract should name the starting date, review dates and owner for each measure. It should also say when a report is due and what records come with it. You need enough detail to trace a number back to the questions or leads behind it. A chart alone does not settle a dispute over what was counted.

When named answers rise, review the question mix alongside qualified leads and source data. Use the same count rules for both periods so your team can see which measure changed.

Agree on what happens when a product changes its interface or a source stops reporting a field. The report should mark the break and explain any replacement. Compare periods only where the definition and scope still match. If the data is missing, show a gap rather than a zero. A zero means a check was made and no qualifying result was found.

Give your team access to the evidence needed for review, subject to your own privacy rules. For lead records, use a safe view that lets the right people check status without sharing more personal data than needed. Name who can correct an error and how the correction appears in the next report. Keep a note of any rule change and its date.

Hold a monthly review as a planning cadence. Check the buyer questions, missing mentions and qualified leads against the agreed rules. Google advises buyers to ask how success will be measured and what results and time frame to expect. Set the review cadence at once a month in the agreement.

What should you put in the contract before signing?

Put each measure in a schedule that both sides can read. Name the result, the source of proof, the count rule, the review date and the person who can answer questions. If a proposal uses a broad term such as visibility, ask for the exact events behind it. The same word can mean a brand name, a link or a visit.

For a local brand, name the brand, question set and AI products in scope. Define a qualifying mention and where the saved records can be found.

State who can approve changes to the question list, the brands in scope and the lead rule. Set a way to resolve a disputed count using the saved record. If a record is missing, say how it will be marked. You can also ask for a sample report before signing, with sample data clearly labeled as such.

Give delivered work and outcomes separate lines. Define the proof for each: a page for a delivery, a saved answer for a mention, and a lead record for a lead.

If a decision or payment depends on results, define the trigger, review period, count source, exclusions and dispute rule. Set a monthly review date and have your legal and finance teams assess the language.

I would want a sample row for each measure alongside the chart. For an answer check, the row can show the question, product, date, observed brand name and link, if present. For a lead, it can show the event, qualification status and source evidence available to the reviewer. A sample makes it easier to spot a vague rule before that rule affects a report. Label the examples so no one mistakes them for actual results.

Walk through an edge case together before signing. If an answer names the brand but links to another site, decide which column gets a count. If a buyer submits a form and later books a call, decide how the record shows both actions without creating another lead. Write the answer next to the measure it governs, so the person preparing the report and the person reviewing it use the same rule.

The agreement should also say what the review can change. A team may correct a data entry, add a new question or revise a lead rule, but those changes have different effects on the comparison. Keep the original period visible and add a dated explanation. When both sides cannot verify a result from the available record, flag it for review instead of silently adding it to the total. Bring disputed entries to the monthly review.

What should your team check before signing?

• Name each brand, market and AI product in scope
• Attach the buyer questions and mark the baseline
• Define a brand mention and a link to your page
• Show the total questions checked beside each count
• Define a lead and a qualified lead in the same system
• Separate tagged visits, buyer reports and unknown sources
• Name the report owner and the review dates
• Set a monthly review and a way to correct records and approve changes

What can these measures tell you and what can they miss?

An answer record captures a particular check, while a source field may leave part of a buyer's path unknown. Use each measure for the event it records.

Keep visits, leads, qualified leads and closed business in separate columns. For revenue, record how your team linked it to the lead and any other sources that played a part.

A reported business result is an example of what one person attributes to the work. It does not set an average or a forecast for your business. Connect answer evidence with records your own team can check.

Use a saved answer to check a mention and a sales record to check a lead. Review both at the monthly meeting, and mark a missing source as unknown.