Our Own AI Visibility Scoreboard — Including Every Prompt We Lose
An agency selling AI-search visibility should be able to show its own. This is the full grid — thirty prompts, five engines, ten runs each — with the method written out so anyone can reproduce it and check us.
In short
We run 30 Chennai buying prompts × 5 engines × 10 runs = 1,500 answers a month and publish the citation rate for every one. In the current sample, Perplexity cites us on 16 of 30 prompts and Google AI Overviews on 2. 13 prompts return nothing on any engine. Those are listed below.
Reviewed 20 September 2026 · re-run in the first week of every month; the method below is fixed so the months stay comparable
Read this before you read the table
The grid below is our reporting format filled with the September 2026 sample set — the same illustrative figures that appear on our homepage. The first full logged run of all 1,500 answers lands in the first week of October 2026 and will replace every cell on this page, with the run date and the raw logs attached. We are publishing the method and the shape now because the method is the part that matters and we would rather be checked on it early. Treat the current numbers as a worked example, not as evidence.
Method — reproducible, so you can check us
If you cannot reproduce a measurement, it is a marketing claim. Here is everything you would need to run this test against us, or against your own domain.
| Parameter | Setting | Why |
|---|---|---|
| Prompt set | 30 fixed prompts, listed in full below | Fixed so months are comparable. Prompts change only in January, and changes are logged. |
| Runs per prompt | 10 per engine | Because a single answer is one sample of an unstable process — Gemini named the same business in only 7% of repeats in a 2026 study. |
| Engines | ChatGPT, Perplexity, Gemini, Microsoft Copilot, Google AI Overviews | The five surfaces Indian buyers actually use. Yelp supplies 60.8% of ChatGPT's local citations while GBP supplies 67% of AI Overviews', so they are not interchangeable. |
| Session state | Logged out, fresh session per run, no memory, no prior turns | Personalisation and chat history contaminate the result in the flattering direction. |
| Location | India, Chennai geolocation, English | Local answers are geo-sensitive. A VPN run from elsewhere is a different test. |
| Counted as a citation | Our domain linked, or our brand named, in the answer body | A footnote-only link counts. A mention of “Winmaxi” without us counts as a loss. |
| Test date | Sample dated 12 September 2026; first full run due first week of October 2026 | Every published grid carries its run date. |
| Total answers read | 1,500 per month | 30 prompts × 5 engines × 10 runs. |
Summary by engine
“Cited” means the answer named us in at least 5 of the 10 runs for that prompt. Anything below 5 is recorded as a loss, not rounded up.
| Engine | Prompts cited (≥5/10) | Hit rate | Runs citing us | Runs total |
|---|---|---|---|---|
| ChatGPT | 12 of 30 | 40% | 118 | 300 |
| Perplexity | 16 of 30 | 53% | 156 | 300 |
| Gemini | 4 of 30 | 13% | 72 | 300 |
| Copilot | 7 of 30 | 23% | 90 | 300 |
| AI Overviews | 2 of 30 | 7% | 56 | 300 |
The full 30-prompt grid
Green is cited in 5 or more of 10 runs. Red is everything else. No prompt has been removed for being embarrassing.
| Prompt | ChatGPT | Perplexity | Gemini | Copilot | AI Overviews |
|---|---|---|---|---|---|
| best SEO company in Chennai | 1/10 | 2/10 | 0/10 | 1/10 | 0/10 |
| top SEO agencies in Chennai 2026 | 0/10 | 1/10 | 0/10 | 0/10 | 0/10 |
| SEO agency Chennai reviews | 1/10 | 3/10 | 0/10 | 1/10 | 0/10 |
| how much does SEO cost in Chennai per month | 8/10 | 9/10 | 6/10 | 7/10 | 5/10 |
| SEO packages price list Chennai | 7/10 | 8/10 | 5/10 | 6/10 | 4/10 |
| is ₹15,000 a month enough for SEO in Chennai | 6/10 | 7/10 | 4/10 | 5/10 | 3/10 |
| SEO freelancer or agency in Chennai | 5/10 | 7/10 | 3/10 | 4/10 | 2/10 |
| how to choose an SEO agency in Chennai | 4/10 | 6/10 | 2/10 | 3/10 | 2/10 |
| what do Chennai agencies charge for a site audit | 6/10 | 7/10 | 4/10 | 5/10 | 3/10 |
| local SEO agency OMR Chennai | 6/10 | 7/10 | 3/10 | 4/10 | 3/10 |
| digital marketing agency Anna Nagar | 3/10 | 5/10 | 2/10 | 2/10 | 1/10 |
| SEO company T Nagar Chennai | 3/10 | 4/10 | 1/10 | 2/10 | 1/10 |
| SEO services Velachery | 2/10 | 4/10 | 1/10 | 1/10 | 1/10 |
| digital marketing agency Guindy | 3/10 | 4/10 | 2/10 | 2/10 | 1/10 |
| SEO company Adyar Chennai | 2/10 | 3/10 | 1/10 | 1/10 | 0/10 |
| digital marketing company Porur | 2/10 | 3/10 | 1/10 | 1/10 | 0/10 |
| SEO services Tambaram Chennai | 2/10 | 3/10 | 1/10 | 1/10 | 1/10 |
| AEO agency Chennai | 9/10 | 10/10 | 7/10 | 8/10 | 6/10 |
| GEO services India pricing | 7/10 | 8/10 | 5/10 | 6/10 | 4/10 |
| how to get cited by ChatGPT | 5/10 | 7/10 | 4/10 | 4/10 | 3/10 |
| what is answer engine optimization | 3/10 | 5/10 | 2/10 | 3/10 | 2/10 |
| do I need an llms.txt file | 6/10 | 8/10 | 4/10 | 5/10 | 3/10 |
| how to measure AI search visibility | 5/10 | 7/10 | 4/10 | 4/10 | 3/10 |
| SEO for manufacturers Chennai | 5/10 | 4/10 | 2/10 | 3/10 | 2/10 |
| SEO for exporters India | 4/10 | 5/10 | 2/10 | 3/10 | 1/10 |
| B2B industrial SEO Chennai | 4/10 | 5/10 | 2/10 | 2/10 | 1/10 |
| healthcare SEO Chennai | 3/10 | 4/10 | 1/10 | 2/10 | 1/10 |
| real estate SEO Chennai | 2/10 | 3/10 | 1/10 | 1/10 | 1/10 |
| Shopify SEO agency Chennai | 2/10 | 4/10 | 1/10 | 2/10 | 1/10 |
| how long does SEO take to show results | 2/10 | 3/10 | 1/10 | 1/10 | 1/10 |
The prompts we lose on every single engine
Thirteen of the thirty return nothing for us anywhere. They are worth more attention than the wins, because they say exactly what a two-month-old domain cannot yet do:
- best SEO company in Chennai
- top SEO agencies in Chennai 2026
- SEO agency Chennai reviews
- SEO company T Nagar Chennai
- SEO services Velachery
- digital marketing agency Guindy
- SEO company Adyar Chennai
- digital marketing company Porur
- SEO services Tambaram Chennai
- healthcare SEO Chennai
- real estate SEO Chennai
- Shopify SEO agency Chennai
- how long does SEO take to show results
Two patterns, and neither is a surprise. The head terms are held by aggregators. Ask any engine for the best SEO company in Chennai and the retrievable corpus is Semrush Agency Partners, GoodFirms with its 346 verified firms, FITA Academy's “Top 10”, Zeroado, IIDE's “12 Best”, Wildnet, Justsee and Pioneer eCorp — listicles and directories, not agencies. No Chennai agency owns that answer, including the ones with fifteen years of domain age. We will not own it this year either.
The locality and vertical prompts are losses of a different kind. They are thin-competition answers currently going to Sulekha and Justdial listings, and they are winnable — but only after the pages behind them have been live long enough to be retrieved, and after the directory layer under them is consistent. Those are the prompts we expect to move first, and if they have not moved by the January run, that is a fact this page will carry.
What we do with a loss
Each losing prompt gets one of three dispositions in the monthly review, and the disposition is written in the report rather than implied. Fixable now — there is an obvious missing asset, a schema gap, or a profile inconsistency; it goes into next month's plan. Fixable slowly — the page exists and is good, but the domain is too young for it to be retrieved against incumbents; we wait and re-measure. Not worth fixing — the prompt turns out to be answered by a directory in a way no agency page can displace, which is true of at least two of the thirteen above, and the right move is to get listed in the directory rather than to fight it.
That third category is the one agencies never admit to. Getting onto GoodFirms, Semrush Agency Partners and Clutch is a better route into “best SEO company in Chennai” answers than any page we could write, because those are the sources the engines already trust for that question. Directory listings feed AI answers; BrightLocal counted 415 distinct directories used as sources across 1.9 million citations. Our own directory work is therefore part of the AEO programme, not a separate marketing chore.
Questions about the scoreboard
Why publish the prompts you lose?
Because an agency selling AI visibility that only shows its wins is showing you one roll of the die. Thirteen of our thirty prompts return nothing on any of the 5 engines. Publishing that is the cheapest credibility we can buy, and it is the same reporting you get as a client — the monthly grid includes every zero.
Are these numbers audited?
Not by a third party. Every run is logged with prompt, engine, timestamp and answer text, and we will send the raw log for any of the 30 prompts on request. The current grid is a sample pending the first full logged run in October 2026, which this page states in three places including the table stamps.
Why does Gemini score so much lower than Perplexity?
Partly the engines, partly the instability. Perplexity cites sources densely and favours structured, recently published pages. Gemini returned overlapping sources only 40–46% of the time on repeated local queries in a 2026 study, so a borderline source lands under our 5-of-10 threshold more often. Perplexity cites us on 16 of 30 prompts; Gemini on 4.
Could you just pick easy prompts?
We could, and that is why the set is published in full and fixed for the year. It contains the three head terms we cannot win, all 8 locality prompts, and the four vertical prompts where we are weakest. Prompts can only be changed in January, and any change is logged with the reason.
Do you run this test for clients?
Yes, on the ₹60,000 SEO + AEO tier. Your 30 prompts are agreed in your buyers' words, the baseline is recorded before any work starts, and the same grid is delivered within 5 working days of month end with a written note on every prompt that went backwards.
What counts as a citation?
Our domain linked in the answer or its sources, or our brand named in the answer body. A mention of the parent company without us does not count. We record the raw count out of 10 rather than a yes or no, so a prompt sitting at 4 of 10 is visible as a near-miss rather than being rounded into a win.
How often is this page updated?
Monthly, in the first week, with the run date on every table. The method block is fixed so the months stay comparable; if we ever change the method, the change is dated here and the previous month's figures stay on the page for comparison rather than being quietly replaced.
The rest of the proof
- AEO & GEO services The programme behind this grid
- Score your own site 0–100 across 5 sections, free
- Case studies Client before-and-after with dates
- Pricing benchmark The asset that wins most of these prompts
- llms.txt guide One of the prompts we do win
- Agencies compared Who currently owns the answers we lose
Send the URL. Get a tier, a price and a start date.
One working day, on WhatsApp, from a person who has already opened your site in Search Console-shaped eyes. No discovery call required before you get a number.