AI SEO: how to actually run the programme

Measure first. Fix later.

You cannot improve a number you do not have. Type your category into ChatGPT, Perplexity, Google AI, and Claude. Ask the same buyer questions a real customer would ask. Who gets named? Where are you missing entirely? Which competitor takes the recommendation when you lose?

Recommendation rate is the metric. It is how often your brand is named when a buyer asks AI engines for advice in your category. Measured across all four engines and three query types (category, comparison, use-case), it tells you what to fix and lets you check what worked after you ship.

• Category queries ("best X for Y") sit at high intent and are the hardest to win
• Comparison queries ("X vs Y") sit at mid-funnel and respond fastest to citation work
• Use-case queries ("how do I do X") sit at top of funnel and respond fastest to content work

Fix entity clarity before anything else.

Entity clarity is the cheapest signal to fix and the one most likely to be broken. If ChatGPT cannot describe your brand the same way twice in a row, neither will the answer it gives buyers.

Look for three breaks. Category drift: do different pages call you different things (a CRM here, a "revenue intelligence platform" there, a "sales OS" on Twitter)? Audience drift: who actually buys from you, and does every surface say the same thing? Name drift: is your brand name spelled, capitalised, and styled the same way everywhere?

Fix them on your homepage, your /about, your /pricing, your social profiles, and your Wikipedia entry if you have one. Same category. Same audience. Same name. Same description.

Publish the pages buyers ask AI.

AI engines do not invent answers. They retrieve sources. A page that directly answers a buyer question is a candidate source. The shortest path to a higher recommendation rate is publishing the pages your category's buyers actually ask about, written so an AI can extract the answer without contortion.

• Category pages: one per high-intent buyer query ("best X for Y"), written as direct prose, not keyword salad
• Comparison pages: head-to-head with two or three real competitors, with honest tradeoffs the model can quote
• Use-case pages: how to do X with your product, end to end
• FAQ pages: short, direct, answer-first paragraphs that match the way buyers phrase the question

Get cited where AI engines actually read.

Your own site is one signal. AI engines weight third-party sources heavily because cross-source consistency is the trust signal. Where the engine reads depends on your category. The list below is India-first, since that is where most AnswerTrace customers operate today; G2 and Capterra are global, the rest are the surfaces Indian buyers and AI engines actually parse.

• B2B SaaS in India: LinkedIn (very high signal), G2, Capterra, YourStory, Inc42, The Ken, Moneycontrol, Entrackr, and product-specific YouTube reviews
• B2C and D2C in India: Amazon.in, Flipkart, Myntra, Nykaa, 1mg; MouthShut, Google Reviews, Justdial; Gadgets360, Smartprix, Beebom, GadgetByte; YouTube and Instagram creator content
• Cross-vertical in India: LinkedIn, Quora, YouTube India, Instagram, and independent comparison pages and listicles for your category

Run the loop. Compound the result.

AI SEO is not a project that finishes. It is a loop. Every quarter, re-measure recommendation rate, look at what moved, and ship the next batch of pages and citations against what you still lose. The discipline compounds: every quarter's answer-ready page and every quarter's new citation builds trust signal that pays off in every subsequent quarter.

AnswerTrace runs that loop in software. Measure recommendation rate across the four engines, diagnose which queries you lose and to whom, surface the highest-impact fixes for your team to ship, and report what changed all the way from AI query to revenue. Start a free measurement run and see your baseline in two minutes.