Can you improve AI search visibility yourself?

Yes—start with a bounded diagnostic, not a monitoring subscription or a promise to make ChatGPT recommend you. The practical DIY route is to identify buying questions, collect repeated observations, separate genuine gaps from measurement problems, improve information you control, and recheck comparable results.

This guide proposes a 90-day programme, preceded by an optional small diagnostic. Its schedules, staffing allowances, prompt counts and decision rules are planning recommendations, not measured industry averages. Published software prices are identified separately. No client-specific assistant testing, supplier product trial or customer interview was conducted for this document.

The central distinction is simple:

• Mention: the answer names your business or product.
• Citation: the answer links to a source. That might be your website or somebody else’s page about you.
• Recommendation: the answer presents your business as a suitable option for the buyer’s requirements.

Use these as separate operational definitions. A negative mention is not a recommendation; a citation to your educational article need not recommend your product.

Your objective is a defensible investment decision: which buying-question gaps deserve work, who can fix them, and whether subsequent observations justify continuing. It is not a single impressive-looking visibility score.

What should you have at the end?

Your DIY programme should produce:

1. An approved commercial hypothesis and spending limit.
2. A versioned buying-prompt library.
3. Raw answers, timestamps and separate mention, citation and recommendation scores.
4. A competitor citation map and product-accuracy register.
5. A prioritised technical and editorial backlog.
6. A deployment log showing exactly what changed.
7. An analytics and customer-discovery report.
8. A continue, change, pause or stop decision.