How we approach GEO

1. 01

   Define the business goal and baseline

   Agree on the buyer, market, decision, and starting point.

   Inputs

   Product materials, target customers, priority markets, competitor shortlist, and available analytics.

   • Choose the decisions to support: product discovery, shortlisting, comparison, migration, or implementation.
   • Agree on languages, platforms, search modes, business conversions, and who can verify product facts.
   • Capture initial answers and existing page coverage before starting the first content changes.

   Output

   Scope sheet, baseline observations, product fact checklist, and measurement definitions.

   Review criteria

   Each goal has a defined audience, a measurable observation, and an owner. Available data and missing access are documented.

2. 02

   Research questions and map answer gaps

   Use Prompt Search to build a question set writers and analysts can share.

   Inputs

   Customer interviews, sales and support questions, search queries, product use cases, and the baseline.

   • Group prompts by buying stage, task, constraints, language, and branded or unbranded intent.
   • Review which brands appear, what is said about them, and which pages are visibly cited in the sampled answers.
   • Prioritize a gap when it matters to the business, lacks a useful answer, and can be addressed with evidence the team can supply.

   Output

   Versioned prompt library, competitor and source map, and a prioritized prompt-to-page backlog.

   Review criteria

   Each priority question has a research basis, an intent label, and a page decision: create, improve, combine, or defer.

3. 03

   Build a reliable evidence foundation

   Establish what the brand can accurately claim and demonstrate.

   Inputs

   Product documentation, approved customer examples, original tests, pricing information, and expert input.

   • Connect material claims to a source, its date, a verification owner, and any conditions or limitations.
   • Use consistent product names, category definitions, feature descriptions, and author or organization information.
   • Identify missing evidence before drafting: request a demo, test, screenshot, interview, or product-owner confirmation.

   Output

   Claim-to-source register, approved brand fact sheet, and a list of evidence gaps to resolve.

   Review criteria

   Quantitative, comparative, and capability claims can be checked. Unsupported claims are removed or rewritten with the appropriate limits.

4. 04

   Produce content around the decision

   Turn the prompt and evidence into a page that answers a real buying question.

   Inputs

   Approved prompt groups, page mapping, evidence register, and editorial brief.

   • Select the right format: a use-case guide, comparison, alternative, how-to, FAQ, or an update to an existing page.
   • Give a clear answer, explain who it fits, show the evidence, state tradeoffs, and provide the next useful action.
   • Review facts, completeness, readability, internal links, and conversion paths before client approval.

   Output

   Approved article, source notes, metadata, internal-link plan, and publishing instructions.

   Review criteria

   The page answers its primary question without requiring a sales call to understand the basics; important statements have support.

5. 05

   Publish, verify access, and distribute

   Connect the content to the site and the relevant places buyers research.

   Inputs

   Approved content, publishing access, website configuration, and relevant source opportunities.

   • Check the published page, indexability, canonical URL, internal links, and whether the main answer is accessible as text.
   • Use applicable structured data that describes the visible content, and route technical changes to the implementation owner.
   • Prioritize relevant editorial, community, resource, or video opportunities found in the source review. Adapt the material to each context and make brand involvement clear.

   Output

   Published URL register, access checks, technical tickets, and records of agreed distribution work.

   Review criteria

   The final page matches the approved facts and can be reached from relevant site pages. Publishing and technical responsibilities are explicit.

6. 06

   Measure, explain, and improve

   Use GEO Monitoring to decide which questions and pages need the next round of work.

   Inputs

   The frozen prompt set, new observations, published URL register, change log, and available analytics.

   • Repeat checks under recorded conditions and separate brand mentions, citations, traffic, and business conversions.
   • Compare like-for-like prompt groups; annotate platform, sample, or content changes and recheck unusual results.
   • Select a small set of changes for the next cycle. Where practical, compare changed pages with stable pages and treat possible causes as hypotheses.

   Output

   Evidence-backed review, prioritized actions, named owners, and a record of what will be checked next.

   Review criteria

   Every recommendation points to an observation and a concrete action. Reporting states sample sizes and coverage changes.

Brand mention rate

Valid answers mentioning the brand ÷ all valid sampled answers

Are we appearing in the selected buying conversations? Compare the same prompt version and platform; record failed checks separately.

Owned-page citation rate

Answers citing an owned URL ÷ valid answers with observable citation status

Is our content used as a visible source? Distinguish a confirmed zero citation from an unavailable citation record.

Answer accuracy

Reviewed descriptions checked against the approved product fact sheet

Is the answer describing the right product, capabilities, and limitations? Log incorrect claims and the pages that need clearer explanations.

Business response

Identifiable referral visits → engaged visits → agreed conversion events

Are identifiable visitors taking a useful next step? Report attribution limits; an AI citation alone is not a visit or a lead.

Reviews include sample counts, failed attempts, content change dates, and platform conditions. We preserve the distinctions between data sources and investigate the evidence behind changes before choosing the next test.

1. Start: agree and diagnose

   Confirm the market, product facts, access, owners, prompt sample, and initial observations. Resolve missing inputs before committing the production queue.

2. First batch: research and produce

   Approve priority briefs and evidence, complete editorial review, and hand the first batch to the publishing owner. The pace follows evidence and review readiness.

3. Weekly: unblock delivery

   Review drafts, approvals, publishing, technical tickets, and evidence gaps. Record decisions and the next actions for both teams.

4. Monthly: review and reprioritize

   Review comparable observations and business signals, decide what to update or create, and revise the next content and monitoring cycle.

• Google Search Central · AI features and your website

  Google keeps foundational SEO requirements for its AI search features and does not require special AI markup. We use this guidance when checking access and page structure.

• Bing Webmaster Tools · AI Performance

  Bing distinguishes citation activity from visits and explains the limits of aggregated grounding-query data. We keep platform reports separate from our sampled prompt observations.

• Aggarwal et al. · GEO: Generative Engine Optimization

  This research frames visibility as something to evaluate and shows that results vary across domains. It informs our testing approach; benchmark gains are not client forecasts.

The engagement commits to the agreed research, content, implementation guidance, and reviews. Platforms generate their own answers; we improve the work through verifiable observations and label conclusions that still need testing.