AI Share of Voice Rose From 2.26% to 12.94%: A Tutoring Case

A specific customer question can be a better starting point than the broad topic a large competitor already dominates. A Canadian tutoring case offers a useful way to think about AI search optimization: helping AI answers understand and introduce a business.

For a local learning centre, that may mean answering questions about the school system its families actually use. This guide takes a content owner from real inquiries to one article topic with a clear purpose.

The content focused on local parents' questions

Genie Teacher connects Canadian families with certified teachers. Agency LoudFace reports building content around Ontario parents' report-card and exam questions and tracking 81 AI prompts. Reported AI share of voice rose from 2.26% on May 25 to 12.94% on August 24, 2026. This is a vendor-specific metric, not revenue or market-wide awareness. Tracking had a gap. It is an agency self-report, not proof of an isolated treatment effect.[1]

The useful exercise is to identify whom an article serves and which question it resolves. The workflow below is our recommendation, not the company's unpublished editorial process.

Open actual inquiries and extract useful questions

Start with recent inquiries or notes from customer conversations. Choose questions about services the business can actually provide. Remove names and addresses, retaining the decision the person needed to make and the missing information.

The table uses a fictional local learning centre. These are not Genie Teacher's customer records or working documents.

Example question	Decision the reader faces	Evidence to gather
My child is falling behind in class	Whether the centre can help with the relevant work	Supported grades, teaching content, staff qualifications
Can lessons fit around extracurricular activities?	Whether a workable schedule exists	Timetable and rescheduling conditions
Can we decline after a trial lesson?	Whether trying the service creates an obligation	Trial fees and enrolment process

Keep the question in recognizable language at first. “Can we just try a lesson?” is more useful for drafting than “enrolment decision barriers” because it names the answer the reader needs.

If inquiries are scarce, collect explanations staff repeatedly give. If those are unavailable too, label proposed questions as editorial hypotheses. Do not present imagined customer concerns as observed conversations.

Group questions by the answer they need

“Can we just try?” and “Can we decline after the trial?” may need the same explanation. Different wording does not require separate articles.

Rescheduling after joining is a different question. Trial conditions do not answer it; it may belong with the explanation of ongoing attendance.

Add a note beside each question: can the same answer resolve this and another question? For each group, inspect the existing site before drafting a new page.

Existing coverage	Editorial action
A page already answers the question	Update outdated conditions or missing details
Answers are scattered across pages	Bring the explanation together and link to details
No answer exists, but evidence is available	Consider a new article
The business lacks evidence or the relevant service	Set the topic aside

Do not create identical articles with different place names. Explain conditions that actually matter in that location. If the school system differs, the official evidence may differ too. Without meaningful local differences, one useful explanation may be enough.

Choose search terms that match the question

Once the questions are organized, investigate the language people use to search. SEO means helping searchers find a relevant page. An article's title should identify what the reader can learn.

For example, “Can tutoring fit around school activities? Check schedules and rescheduling” says more than “Optimal learning solutions.” Use such a title only if the article actually explains those conditions.

A long-tail keyword describes a more specific subject or need. Adding more words does not guarantee a ranking. “Tutoring around school activities” simply identifies a situation more precisely than “tutoring.”

Large search demand is not a reason to choose a question the business cannot answer. Smaller demand can still matter when it reflects a real decision before enrolment. If a tool returns no volume, record the demand as unknown, not zero.

Google's guidance emphasizes helpful content and demonstrated knowledge or experience.[2] Use that principle to fill a missing answer, rather than treating an extra published page as the goal.

Build one article around an answer

Create a draft in the editor your business already uses. Screens vary, so the table describes what the body needs. If a row cannot be completed, gather its missing evidence first.

Reading order	Example content for a learning centre
Direct answer	Conditions under which lessons can fit around activities
Things to check	Lesson days, journey home, homework, rescheduling
Concrete example	A clearly fictional weekly schedule
When it will not work	Actual timetable constraints
Next action	View the timetable or check trial conditions

Label an invented student or schedule as an example. Use the business's real rules and timetable. Do not invent an improvement in grades or a customer success story.

Ask a colleague who did not write the draft to explain when a reader could enrol and what still needs checking. Their answers help locate missing information. A clear next task for the reader is also a useful test of the article's purpose.

Observe AI inclusion and enrolment separately

Before publishing, fix the questions you plan to test. A request for tutoring compatible with extracurricular activities needs the region and grade appropriate to actual customers. Keep questions that name the business separate from discovery questions that do not.

For continuing observations in AgentSignal, use the AI citation guide to register the site and questions. Login and analysis credits are required. Keep the question, AI service, and other conditions consistent, and inspect both the answer and cited pages.

The source case's share-of-voice measure need not use the same calculation as AgentSignal's citation rate. Its 12.94% is not a target score for a different business or tool.

Record Google impressions and clicks, identifiable AI referrals, and trial applications separately. More AI mentions may not produce more visits. If applications increase, also check for advertising or seasonal changes.

Google says established SEO practices remain relevant to its AI features, including useful internal links and accessible text.[3] Establish those conditions and then observe what changes.

The first article does not need to explain all of education. It needs to answer one problem a customer faces now. Open actual inquiries, check existing coverage, and add a supported answer. Repeating that process builds information people can use to make decisions, alongside the article count.