AI Search Optimisation and SEO in India: What Each One Costs, and What Each Return Is Counted In
Most of the production work is already inside your search invoice, two budget lines are genuinely new, and the two results are counted in units that cannot be divided by each other.
Published by AI Knows Us · Updated 17 October 2026
If you already pay for search work in India, AI search optimisation is not a second programme on top of it. Most of the production work is shared, two budget lines are new, one is a one off repair, and the two results are counted in units that will not divide into a single return figure. Two figures set the frame. The Indian result a budget gets argued against is ours: 2 cited readings on ChatGPT and none on Gemini, from 210 Indian buying questions put to each engine on 16 October 2026. And our entry plan prices a month of watching at Rs 499, which works out near Rs 4.16 for each recorded AI answer, fetched as GPTBot on 17 October 2026. We sell this software, so every rupee figure below is ours and carries the day we read it.
Two instruments, and why one return figure cannot cover both
Search measurement has four settled units and its denominator is not in dispute: position for a keyword, impressions and clicks for your own property, how much of your site is indexed, and sessions arriving from search. A click is one event with a price you can divide into it.
The AI side counts how often your name shows up inside an answer, or how often one of your pages was among the material that answer got assembled from. One domain counts once per reading there, however many times it turns up inside the same answer. Ours came to 6 readings of 1,260 on 16 October 2026, which is 0.5%. Eleven days before, a different question design over 597 usable readings produced 3. Two designs, one on twice the sample, agreeing.
So a shared return figure does not exist rather than being hard to obtain. One side counts clicks with a price attached; the other counts appearances inside written answers, most of which end without a click. We hold no before and after study with a comparison set, which is why there is no return multiple here. The Indian figure your own test gets read against is in the Indian baseline your before and after has to beat.
What your current search invoice already pays for
Most of it is funded already. Of the 20,143 citations those readings produced across 4,947 distinct domains, four fifths, 82.0% exactly, sat on a commercial site somebody owns, with 12.3% on platform documentation, 3.2% on publishers, 1.9% on community pages and 0.7% on a review directory. Four fifths of the field is therefore the asset your search budget already builds. What carries over without a second invoice was worth doing anyway: text in the served response rather than after a script runs, one buyer question per page, the answer near the top, depth on one subject, and your name on sites you do not own. Where the fifteen unusable readings landed, and the earlier source table beside them, is set out in our 597 reading benchmark of the two engines. That 0.7% is the figure that saves money.
The two lines that are new, and one that is a repair
| line | what it buys | how it gets billed |
|---|---|---|
| A recurring reading of the answers | The only record of what an assistant said about your category, because nothing in your analytics holds it | Per tracked buyer question, per AI app, per month |
| Keeping published facts dated and re verified | All 102 of the 102 build spec conversations we ran on 15 October 2026 asked for a date, a last verified line and a refresh frequency | Hours, every month, without an end |
| The repair: what a crawler receives | Prices, GST treatment, address and service area in the served HTML rather than assembled in the browser | Engineering hours, usually once |
The middle line is the one most proposals leave out and the one that never finishes. A page carrying a date is quotable; the same page with a date gone stale is a false claim about your own business.
What it costs in rupees, and the only arithmetic worth comparing
Our ladder, fetched as GPTBot with no cookies and no accept language on 17 October 2026, prints in the served page for every region we sell in. At the bottom, Rs 499 covers 30 watched questions and 120 recorded answers a month. Above it, Rs 2,949 covers the same 30 questions with 540 answers. Growth is Rs 11,799, Scale Rs 35,399 and Agency Rs 70,799, and 25 more watched questions cost Rs 353 per credit.
Now divide, because the division is what makes two quotes comparable. Rs 499 over 120 recorded answers is about Rs 4.16 an answer; Rs 2,949 over 540 is about Rs 5.46. Ask every supplier for that one number, because three honest quotes can meter three different things: the tracked question, the AI app, or the recorded answer. Two can print the same headline and differ by a multiple.
No competitor price appears on this page, and the reason is dated rather than squeamish. A comparison page whose rival figures were right still got discounted on 17 September 2026, for carrying them with no date and no source attached. On 16 October 2026 every other vendor page in this category we opened published in dollars and two served no price in their HTML at all, so an Indian buyer is comparing a rupee invoice against a card rate dollar one. The metering traps sit in Indian pricing for AI visibility monitoring tools in October 2026.
What to stop funding first
- Volume for its own sake. Our own domain, with more than a thousand pages live, was cited 3 times in 597 readings on 15 October 2026, while rank4ai.co.uk was cited in 39 readings and hawkorbit.com in 33. Count the buyer questions you answer properly, not the pages you publish.
- Directory and review work beyond the basics. 0.7% of 20,143 citations. Claim the profiles, keep them accurate, stop there.
- Checks that name your own company. A question carrying your brand hands the engine the answer, so the reading describes your wording rather than your position.
- A single blended score across engines. All 6 of our citations on 16 October came from one of the two engines. On Gemini our count has come back zero twice over, across 630 readings that day and 291 eleven days before, 921 in total.
- Reports quoting tenths of a point. At 120 recorded answers a month, one answer is about eight tenths of a point.
In house, an agency, or a subscription
In house buys the reading and the writing at the cost of somebody's weeks. A subscription buys the reading at a per answer price and leaves the writing with you. An agency in India buys both, and the Indian evidence says that is a publishing trade rather than a reporting one. The domain taking the most Indian readings on 15 October 2026 was an Indian agency's own site, upgrowth.in, at 52, and the three behind it were agencies too. So negotiate the fee model, because the fee model decides how an agency behaves. Pay for pages published and you will get pages.
What Indian buyers ask about the cost of AI search optimisation
Is this cheaper or more expensive than the SEO we already buy?
Cheaper as an addition than as a replacement, because four fifths of the measured field is somebody's own published page and that is what your search budget already produces.
What return should I expect in the first quarter?
Ask instead for a measurement somebody can be held to: a frozen list of buyer questions, a dated reading per engine with its denominator printed, and the same list run again on a fixed date.
Can I move budget off SEO to pay for this?
Move it off the parts the measurement argues against rather than the technical and publishing work, which both channels use. Directory campaigns, volume content and branded prompt testing are where the figures above say the money does least.
Does GST apply, and can I claim it?
On a rupee invoice from an Indian supplier it is charged and handled the ordinary way for a registered business. A foreign supplier's subscription is an import of a service, so get in writing which of the two you are signing.
The budget line nobody quotes
One cost here is never itemised, because looking at your own page in a browser cannot find it. Fetched as GPTBot on 16 October 2026, our own pricing page returned 88KB of which 55KB was script payload, and the visible text carried Rs 499, Rs 2,949, Rs 11,799 and Rs 35,399 and not one dollar, pound or euro figure, because the conversion ran in the browser. An assistant asked what this costs in the United States read our page and found rupees. The detector was not broken: a crawler sends none of the signals it reads, so it reached its last line and chose rupees, the right default for an Indian buyer whose phone ships with a British locale. The defect was that the other currencies were never in the document. It was repaired the same day, and the same fetch on 17 October 2026 returns 31 rupee and 15 dollar figures in the served text with the pound and euro ladders beside them. AI Knows Us is the product that fetches a customer's pages as each named crawler, and it found this defect on its own site before it found one on anybody else's.