Every form on this website writes one row to one Google Sheet. On 17 November 2026 we read all 36 rows written since the first on 29 May, 171 days, and sorted them. Eleven were a business asking for help. Eight were us testing our own forms. Four were somebody pretending to be a consumer brand. Four were junk from the first week of a Google Ads campaign. The remaining nine were a pricing tool used three times by one visitor, three vendors pitching us, one job seeker, one gibberish row and one pin code with nothing else. This is the complete funnel of a small agency's website, published because most agencies publish the eleven and not the twenty five.
A word on the words. Search engine optimisation, SEO for short, is showing up on Google without paying per click. AI search optimisation, AIO on this site, is being the business ChatGPT or Gemini names when a buyer asks. The sheet is where both are supposed to end up, so it is the honest place to measure them.
The thirty six rows
| What the row was | Rows | Senders | Pages they came from |
|---|---|---|---|
| A business asking for help | 11 | 10 | Clinic page 4, manufacturing page 2, home page 2, studio, contact, one blog post |
| Our own tests | 8 | 4 | Doctor, clinic, B2B, pricing, UK, education, services |
| Brand impersonation from a lookalike domain | 4 | 3 brands | Home page |
| Junk through a Google Ads click | 4 | 2 | Services page |
| Pricing tool, self serve | 3 | 1 | Pricing page |
| Vendors pitching us | 3 | 3 | Contact page |
| Job seeker, gibberish, a pin code alone | 3 | 3 | Careers, contact, clinic page |
The eleven enquiries came from ten businesses, because one American company wrote twice, on 11 and 16 November, with the same message, five days apart, having had no reply to the first. That row is the most expensive on the sheet and we come back to it.
Where the eleven came from
Seven came through two Meta campaigns that ran on the clinic and manufacturing pages between 29 May and 18 June: four clinics, a general physician among them, and two manufacturers, one of office furniture and one of pipe. The seventh Meta row carried a six digit pin code and nothing else, so we count it as a fill and not as an enquiry, which makes the honest Meta count six. The campaigns stopped on 18 June and no clinic or manufacturer has written through a form since.
One came on 29 August from a blog post about marketing for interior designers, and the visit carried a referral from ChatGPT. We wrote about that enquiry in September; it remains the only organic form fill from a Hyderabad business in the whole sheet, and it came from an article, not from a service page.
One came from the studio page in June, asking about explainer videos. Three came from the United States: a published author reopening a video channel, through the contact page on 25 October, and a company wanting Google Ads and Meta Ads for brand awareness and leads, through the home page on 11 November and again on 16 November. Since 1 October every genuine form fill has come from outside India. Zero have come from Hyderabad.
The original number. All 36 rows of the website form sheet, 29 May to 16 November 2026, read in full on 17 November and classified: 11 enquiries from 10 businesses (7 through two Meta campaigns 29 May to 18 June, of which 1 carried only a pin code; 1 from a blog post with a ChatGPT referral on 29 August; 1 from the studio page; 3 from the United States, 25 October to 16 November, 2 of them one sender); 8 internal tests from 4 of our own numbers; 4 brand impersonations from lookalike domains of 3 consumer brands (1 July, 17 October, 20 October, 28 October); 4 junk rows through a Google Ads click (13 and 14 November, 3 of them one sender in four seconds); 3 pricing tool self serve rows from 1 visitor; 3 vendor pitches; 1 job seeker; 1 gibberish row. Genuine share 31 percent of all rows, 39 percent of rows that were not our own tests. Organic enquiries from a service page: 0. Logged as data ledger entry 86.
The twenty five
Eight rows are us. Two in June from the doctor and clinic pages, one in August from the B2B page, one from the pricing tool, one from the UK film request form in September, two from the education page in October and November, and one from the services page on 16 November, sent on purpose to check whether the sheet still received what the site sent. It did, within the minute. A sheet that is 22 percent test rows is not embarrassing; a sheet nobody tested would be.
Four rows are impersonations. Each claims to be the marketing team of a consumer brand, two of kitchen appliances and one of fashion, exploring an agency partnership. Each arrives from a domain that resembles the brand's and is not it. Each carries a phone number from another country pasted behind an Indian prefix, because our form added the prefix for them. Two carry the same text from two different lookalike domains three days apart. They are the best written rows on the sheet, and the only ones that address us by what we do.
Four rows are what the first week of a Google Ads campaign produces when the ad sends people to a services page and the form asks three questions. One is a four letter string and an emoji at eleven at night. Three are one surname, sent three times in four seconds because the button accepted repeated clicks; we added a guard the next day. All four carry the campaign id in the page address and nothing in the campaign column, because the ad passes a click id and the form was built for UTM tags. Four fills, zero enquiries, and the only four rows that tell us an advert was running.
Three rows are one visitor using the pricing tool three times over two weeks in August and reaching the self serve tier each time. Three are vendors: a social media freelancer, a writer compiling a hundred agency listicle, a follow up about an article on edge AI. One is a job seeker who used the careers form correctly. One is gibberish from a disposable domain. One is the pin code.
What the sheet says that the analytics did not
Three things, in the order they cost us.
First, the second American message. A company asked for ads help on 11 November, waited five days, and asked again. The sheet received both rows within the minute. Nobody read them, because the sheet's notification was not switched on and we had been reading the sheet once a day from a script that only reported counts to a status file. The fix is an email on every row, which is a three minute setting, and we had not made it in 171 days. Every form that works and nobody watches is a form that does not work.
Second, the source column. The UTM fields filled in for the Meta campaigns and the ChatGPT referral, and sat empty for the Google Ads rows, so the one paid channel running this month is invisible in the column built to show paid channels. The page address carries the campaign id; the form should copy it across.
Third, the geography. The pages were written for Hyderabad and the three genuine enquiries since October are American. The October enquiries post counted twelve attributable enquiries and found four of them by telephone and none from an organic visit to a service page; a month later the form sheet agrees, and adds that the organic visitors who do write are not in the city the pages name. That is either a page problem or a market signal, and the Marketing Operating System exists so that we decide which with the next month's rows rather than argue it now.
Frequently asked questions
How many of a small agency's form fills are real enquiries?
On our sheet, 11 of 36 between 29 May and 16 November 2026, from 10 businesses, which is 31 percent of all rows and 39 percent once our own 8 test submissions are removed. The rest were 4 impersonations of consumer brands from lookalike domains, 4 junk rows from a Google Ads campaign in its first week, 3 self serve submissions from one visitor to the pricing tool, 3 vendor pitches, 1 gibberish row, 1 job seeker and 1 row carrying only a pin code. One sheet is not an industry benchmark, but it is a complete one, and most agencies publish the 11 without the 25.
Where did the genuine enquiries come from?
Seven of the 11 came through two Meta campaigns on the clinic and manufacturing pages between 29 May and 18 June. One came from a blog post about interior design on 29 August, the visit carrying a ChatGPT referral. One came from the studio page in June. Two came from the United States through the home page, from the same sender on 11 and 16 November, and a third American wrote through the contact page on 25 October. Since 1 October every genuine form fill has come from outside India, and none from Hyderabad.
What is a brand impersonation row?
A message claiming to come from a well known consumer brand's marketing team, offering an agency partnership, sent from a domain that resembles the brand's but is not it, with a phone number from another country pasted behind an Indian prefix. We received four, from three brands, between 1 July and 28 October 2026, two of them the same text from two different lookalike domains three days apart. They are the most professional looking rows on the sheet, which is the point of them.
What does a form sheet have to do with AI search optimisation?
Search engine optimisation, SEO, is showing up on Google without paying per click. AI search optimisation, AIO on this site, is being the business an AI tool names when a buyer asks. The sheet is where both are supposed to end, and it recorded no enquiry from an organic visit to any service page in 171 days. The one organic enquiry came from a blog post, through a ChatGPT referral. The pages built to be found produced tests, pitches and ad junk; the page written to be read produced a customer.