A practical, week-by-week programme for making answer engines name your business — built from what we observed in our own citation audit, not from theory.
- The strategic problem. Try this tonight.
- The operating choice. **Days 8–21, continued: make your facts consistent everywhere.** AI models build a picture of your business from every mention they can find — your website, Google Business Profile, directories, social bios, press.
- The better model. **Days 71–90: re-measure and double down.** Re-run your ten baseline questions across the three engines.
The strategic problem
Try this tonight. Open ChatGPT, Gemini and Perplexity and ask each one: 'best [your category] in [your city]'. If your business is not in the answer, you are invisible to the fastest-growing research channel in India — and one of your competitors is getting recommended instead. This article is the exact programme we run to fix that, laid out over 90 days.
**Why this is winnable right now.** In our citation audit of Indian marketing queries, we found that answer engines lean on a small, repeatable set of signals — and almost no Indian business optimises for them deliberately. The window where early movers get cited by default is still open in most categories and cities. It will not stay open.
**Days 1–7: measure before you touch anything.** Write down the ten questions a real buyer would ask an AI before hiring someone like you — 'best interior designer in Pune', 'how much does a root canal cost in Ranchi', 'which CA firm handles startup compliance'. Run all ten through ChatGPT, Gemini and Perplexity. Screenshot who gets named and which sources get cited. This is your baseline. Everything below is measured against it.
**Days 8–21: fix your own pages so AI can quote them.** Answer engines cite pages that state an answer plainly near the top, use headings that mirror the question, and contain concrete specifics — numbers, ranges, named steps. Rewrite your key pages so each one answers one buyer question in its first fifty words. Add a short FAQ section with real questions and direct answers. Publish your pricing or at least your pricing logic — 'it depends' is a sentence no AI can quote.
The operating choice
**Days 8–21, continued: make your facts consistent everywhere.** AI models build a picture of your business from every mention they can find — your website, Google Business Profile, directories, social bios, press. If your name, services, city and phone number differ across these, the model's confidence drops and it recommends someone more consistent. One name, one description, one set of facts, everywhere.
**Days 22–45: get described by third parties.** For 'best X' questions, answer engines lean on listicles, directories and comparison pages more than on your own site. Identify the lists and directories that already appear in answers for your category — then get yourself accurately described in them. One accurate, detailed third-party mention is worth more than ten pages on your own domain for these queries.
**Days 22–45, continued: publish one thing worth citing.** The single strongest GEO asset is original data — a benchmark, a price index, a small study. Answer engines prefer citing a primary source over quoting someone quoting someone. Our own ad-benchmarks study gets cited because the numbers exist nowhere else. Find the number in your business that nobody publishes, and publish it.
**Days 46–70: answer the long tail in public.** Buyers ask AI the questions they are embarrassed to ask you — cost, red flags, 'is it worth it', 'what goes wrong'. Each of those is a page. Write them honestly, including the uncomfortable parts. A page titled 'what goes wrong with [your service]' that answers fairly gets cited far more often than a page of superlatives, because it is the only page actually answering the question.
The better model
**Days 71–90: re-measure and double down.** Re-run your ten baseline questions across the three engines. Note where you now appear, where competitors still do, and which sources the engines cite for those answers. Then feed what you learned back into the next cycle: more of the page types that got cited, more presence in the third-party sources that keep appearing.
**What does not work.** Keyword-stuffing AI-facing text into pages, publishing volume for its own sake, and 'AI-optimised' content written by AI about nothing. Our audit found no correlation between publishing volume and citation — a single precise page routinely beats twenty vague ones. The engines are looking for the clearest answer, not the loudest site.
**The honest caveat.** Answer engines are non-deterministic, personalise results, and change models without notice. Nobody can guarantee a citation, and anyone who does is selling something. What this programme does is stack every known signal in your favour — and in most Indian categories today, that is enough to be the only serious candidate.
**If you want the short version.** Answer one question per page, plainly, in the first fifty words. Keep your facts identical everywhere. Get described accurately by third parties. Publish one number nobody else publishes. Repeat for 90 days, then measure again. That is the whole game — and almost none of your competitors are playing it yet.
"A single precise page routinely beats twenty vague ones — the engines want the clearest answer, not the loudest site."
Founder · Strategy & Positioning. Brand strategist leading positioning and category work for Indian and cross-border brands.
Turn this article into a working growth decision.
Bring us your current numbers, website, and agency reports. We'll show you what is signal, what is theatre, and what deserves budget next.
Book an honest audit →