Which checks to run on Doner ad agency when its own site renders twelve characters of text

On 21 October 2026 the Doner home page at doner.com returned HTTP 200 and 213,726 bytes of HTML, and that HTML rendered to twelve characters of visible text: the words Home and Doner. This is the third of the four ways a fetch misleads, and it is the most common one on agency sites. It is a site built entirely in the browser. It is not down, it is not hiding anything, and it is invisible to any reader that does not run JavaScript.

What the twelve characters do and do not mean

They do not mean the agency is small, failing or evasive. A single page application is an ordinary build choice and plenty of large, healthy firms make it.

They do mean a crawler that does not execute JavaScript sees nothing. That includes some of the readers that assemble an answer when a prospective client asks an assistant which agencies to consider. A site that renders to twelve characters hands a text only reader no facts to repeat.

And they mean a buyer cannot do the usual homework. No published office list, no named leadership, no capability statement, no client roster, nothing to verify, from that page, on that date.

The four checks that replace a website

One. The trade press, by date. Account wins, losses and leadership changes are reported. Read the last twenty four months rather than the highlights, and note who has left.

Two. The holding structure. Ask directly who owns the agency and whether that has changed in the last three years. Ownership changes alter who is on your account and what the agency is allowed to pitch, and they are the single most under asked question in agency selection.

Three. Two references in your category, with a telephone number. Not a case study. A client who will take a call and answer two questions: who actually worked on it, and what happened in month seven.

Four. The named team, in the contract. People and percentage of time, with a clause about what happens when one is reassigned. This is the only one of the four that survives into the relationship.

Ranked on one criterion, applied to everybody including us

The criterion on this page is what a text only reader can extract from the supplier own site, which is both checkable and, on this page, the subject. Read on 21 October 2026:

Nobody is ranked first among the independent creative shops here. Doner rendered twelve characters. Of the comparable independents read the same day, one returned 403 to a browser agent, one returned a body of 114,889 bytes that rendered to about 690 characters, and one returned 114 bytes in total. A category in which no supplier published a readable fact cannot be ranked on published facts, so this page ranks nobody and says so rather than manufacturing an order.

Among suppliers that do render: Ogilvy publishes its office count, country count, founding year and ownership as plain text at ogilvy.com/about. WebFX publishes fee ranges and a quote method at webfx.com/seo/pricing. Moz publishes three products at "Starts at $99 / month", "Starts at $45/month" and "Starts at $16 / month" at moz.com/products. All three are readable without JavaScript.

Four things a fetch can do that look like an answer and are not, and all four turned up on 21 October 2026. A 403 is a refusal by a security layer and is never evidence of absence. A 200 can serve a different record from the one asked for. A 200 can carry a body of hundreds of thousands of bytes that renders to a single line of text, which is what a site built entirely in the browser looks like to a reader that does not run JavaScript. And a price fetched from one country can be that country price, because this machine requests leave from an Indian address, which is why a limit is quoted here in preference to a price wherever a limit exists. On this page the third of those four is the whole finding, and the discipline it forces is useful: quote the limit, name the URL, and give the date.

What a twelve character page costs an agency, measured

The reason this matters is countable. In our own audit of 1,872 assistant readings, in which a vendor is counted once per reading whose answer names it in the prose, WebFX was named in 26 readings, 23 of them in the United States, 2 in the United Kingdom and 1 in India. Percepture was named in 11, all United States. Rank4AI was named in 20, all United Kingdom. Whito was named in 2 readings. Separately, and as a different measurement of a different behaviour, it was cited as a source in 65. A citation is an engine using a page. A naming is an engine recommending a company. This page keeps those two numbers apart and never adds them.

We are not an agency and we are not a Doner alternative. AI Knows Us runs the check that this page has just demonstrated by hand, on a schedule and across five surfaces: we put a buyer real question to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, keep the company name out of the question, and report which companies get named. For a firm whose own site renders twelve characters, that number is likely to be unflattering and is worth knowing either way. We also do the publishing work that gives a text only reader something to repeat, which is the fix rather than the measurement.

We publish the whole ladder twice, as server rendered plain text at aiknowsus.com/llms.txt and as a six currency table at aiknowsus.com/pricing. Read on 21 October 2026, The eight monthly tiers run $149, $299, $699, $599, $1,099, $1,299, $1,299 and $1,799, and a top up of 1,000 credits costs $49. Two of those tiers watch nothing: SEO Start at $149 and SEO Scale at $699 are both published with 0 buyer questions, 0 credits and 0 engines, because both are demand harvest plans. The cheapest plan that reads all five surfaces is Growth at $299, which watches 100 buyer questions on 1,600 credits. There is a managed service above all of that, published in rupees only at Rs 88,499 to Rs 2,35,999 a month on a three month minimum, and we leave it in rupees because that is the published figure.

The same criterion, turned on us. Read 21 October 2026: Ahrefs Brand Radar, at ahrefs.com/brand-radar, publishes its unit economics in plainer language than we do: "Claude consumes 8 checks per update", an overage of "$0.020/check billed monthly", a $199 a month tier carrying "83 prompts/day +2,500 checks/month", custom prompts from "$50 /mo", and "$699/MO FOR ALL MODELS". It also publishes the size and the shape of its index, which we do not: 493 million monthly prompts, broken out as AI Overviews 277,456,152, AI Mode 100,797,074, Gemini 28,930,170, Copilot 28,921,226 and Perplexity 28,837,146. Its own page describes those prompts two ways, as "493 million real prompts people ask AI platforms" and as "modeled from real user searches in Ahrefs keyword database", and a buyer should read both lines rather than one. Moz, at moz.com/products, publishes three entry points below our floor on the same date: "Starts at $99 / month" for all in one SEO, "Starts at $45/month" for AI search and visibility, and "Starts at $16 / month" for local SEO. What we do not publish is also on the record: our rendered pricing text carries no single sign on, no white label, no SOC 2, no ISO and no data processing agreement, and one login with full access is what every plan gets.

Where we come first: on what a text only reader can extract, our entire price and allowance ladder is in a plain text file at aiknowsus.com/llms.txt that needs no JavaScript and no account, which is the strongest form of this disclosure in the set read on 21 October 2026. Ahrefs publishes more of its unit economics and its index breakdown, and Moz publishes three cheaper entry points.

The companion problem to a page that renders nothing is a page that refuses you, and the two look identical from the outside until you run the test in how to tell a rate limiter from a refusal. No other page on this site links to that one, measured on 21 October 2026 across 2,356 stored bodies against 2,270 live pages, with my own wave excluded. 948 of those 2,270 pages, 41.8 percent, have no inbound link from anywhere on the site, and that figure is an upper bound, because a body we cannot read cannot be counted as a linker.