Chennai marketing agencies: a source linked directory and pricing dataset

The dated figure that makes this directory usable is the count of named Chennai agencies with public prices and scope verified from a primary page. As of 29 September 2026 that count is 0. We have verified no Chennai agency from a primary page, so this page names none and publishes no prices. What it gives you instead is the source linking standard: the five things a source link must contain before a row counts, how to archive it so the row survives the page being rewritten, and the field list for a dataset anybody can check. The standard exists because of a dated result in another sector. On 17 September 2026 Claude read two comparison pages in legal technology whose competitor prices carried no link, no date and no source. It verified the figures independently, found them correct, and still treated the pages as advocacy and used official sources instead. The page was right and it still lost, because a reader could not tell.

Source linking is the cheapest credibility there is. It also takes discipline, because the temptation on a busy afternoon is to write the number down and move on, and a number without its link is unrecoverable a month later when somebody asks where it came from.

What the directory covers

• Chennai means a publicly stated office address within Greater Chennai Corporation limits. Areas where agency offices are commonly found include T Nagar, Guindy, Nungambakkam and the Old Mahabalipuram Road corridor. Addresses in Chengalpattu or Kanchipuram districts are recorded separately and labelled.
• Agency means a firm that publicly sells marketing services. Named services counted separately: search optimisation, paid media, content, social media, generative engine optimisation.
• Verified from a primary page means the firm's own website page or a public filing, captured with a source link that meets the five part standard below.
• A price counts only when it is a number with a currency on a freely reachable page, read on a recorded date. Prices behind a form are recorded as gated and counted separately.
• Out of scope: firms with no stated address, firms whose only presence is a directory profile, and anything told to us verbally.

How agencies were found and checked

The five part source link standard. A row without all five is not published as verified. This is the whole method and it takes about a minute per field once it is a habit.

• One. The full URL, not the domain, and not a search result link.
• Two. The date you read it, to the day.
• Three. The exact location on the page, such as the section heading or the table the number sits in, so the next reader can find it even after a redesign moves it.
• Four. An archived copy, either a screenshot saved with the row's identifier in the file name, or a public web archive snapshot link. Both is better.
• Five. The claim in the firm's own words, quoted short, so a reader can see whether your summary changed the meaning.

The finding and checking steps.

• One. Build candidates from named sources and record the source per candidate. Search results for service plus city, Google Business Profile listings, the Ministry of Corporate Affairs register, industry association member lists, and the supplier pages of local businesses. A candidate source is not a verification source, and keeping the two columns apart is what makes the dataset honest.
• Two. Freeze the named list with its date before any price is looked at.
• Three. Deduplicate on registered company name, directors, address, and phone or email domain, so the denominator counts firms rather than brands.
• Four. Verify address, services, scope and price from primary pages only, each with a five part source link.
• Five. Archive every primary page immediately. Agency sites are rewritten without notice, and a row whose source has vanished is an assertion again.
• Six. Fetch each pricing page without running scripts and record whether the price is present in what a crawler receives. On 6 August 2026, in legal technology, a pricing page rendered its prices only after scripts ran, so crawlers received no prices at all, and the prices assistants quoted came from an app store listing instead of the company's own site.
• Seven. Publish the counts with the cutoff date, and publish the rows rather than a summary.
• Eight. Log every correction with its date and the previous value, so the dataset has a history rather than a current state.

Named agency profiles

Agencies verified from a primary page: 0. Agencies with a public price verified from a primary page: 0. Cutoff date: 29 September 2026. Nothing is named here. A table of Chennai agencies assembled from other directories would fail part one of the source link standard on every row.

The dated evidence we hold, sectors named, none of it about Chennai or about agencies.

• 0 of 6 questions where any third party review or directory source reached an answer. Legal technology, clawlaw.in, 17 September 2026, Claude. One well known review site appeared in the raw results and was discarded, because the list it offered was of American products and so was not an answer to an India question. If you are counting on a national directory to represent you, that is the risk.
• Official pages and vendors' own pages dominated the citations. AI visibility tooling, aiknowsus.com, September 2026, Perplexity, counted across 24 captures, with a single well known review site far down the list.
• Content used, attribution dropped. Legal technology, clawlaw.in, 17 September 2026, Claude. The assistant admitted it had used two specific arguments drawn from the company's pages and had dropped the attribution in both places, describing it as a citation lapse rather than a ranking judgement. Source linking your own claims will not stop that, and it does give you something to point at when it happens.
• Our own count was 0 of 6. AI visibility tooling, aiknowsus.com, September 2026, Perplexity, six blind questions in our own category, with 161 recorded statements across 24 batches and 72 conversations that we were not cited.

Comparisons by business type

No ranking is published, because no firm has been verified. Six business types, and what to require from a shortlist for each.

• A manufacturer selling to industrial buyers. Require a written scope per service and ask for a sample monthly report. Long sales cycles mean you will judge the work on leading indicators, so the indicators must be defined in month one.
• A healthcare or clinical services provider. Require the firm to state what it will not claim on your behalf. Regulated claims are the fastest way to a problem, and a firm that has never thought about it will not start on your account.
• A software company selling abroad. Require per country reporting and separate engine reporting, and treat one blended global number as unusable.
• A consumer brand on daily paid media. Require the fee shape, the spend assumption and the resulting monthly figure in one table before you compare anything.
• Anyone buying generative engine optimisation. Require the frozen prompt set, check your brand is in none of it, and require the run count arithmetic. On 27 July 2026 the same 78 questions run twice on one day for clawlaw.in returned first place on almost every question where the wrapper named the brand, and second by breadth when nothing was named. The first run was discarded.
• Anyone whose last agency reported well and moved nothing. Require raw captures monthly in your own storage as a contract term, and make the first deliverable a dated baseline rather than a strategy document.

Price dataset and limitations

The dataset is empty on 29 September 2026. Firms named: 0. Verified from a primary page: 0. Public price: 0. Gated: 0. None: 0. Build your own with these ten fields, one row per firm and service.

• Firm name, registered name, and brand names used.
• Address and the municipal area it falls in.
• Service, in the firm's own words.
• Price and currency, or nil.
• Price shape, fixed, starting from, band, hourly, per project, percentage of spend, or tier.
• Scope quoted short, in the firm's words.
• Full URL.
• Date read.
• Location on the page.
• Archive reference, screenshot file name or snapshot link.

Four limitations of this dataset shape, stated so a reader does not over read it.

• It records what firms publish, which is a business decision rather than a measure of quality.
• A published price is a starting position. The dataset cannot see negotiation, and most final prices differ from the published one.
• Rows decay. An unrechecked row older than about three months should be treated as historical, which is why the read date sits in the row rather than at the top of the page.
• An archive is not consent. Screenshot for your own evidence, and where you publish a quote keep it short and attributed to the firm's own page.

Corrections and update log

Corrections. If you run a Chennai agency and want to be verified under this standard, or you can show that something here is wrong, write through the contact page. Every correction is logged with its date and the previous value, and we do not edit a claim away silently.

Sources for every figure above. The clawlaw.in programme of July to September 2026, recorded in GEO_BASELINE_RESULTS_2026-07-27.md and the assistant audit files of 17 September 2026, and the aiknowsus.com audit of September 2026 across 24 batches and 72 conversations. Counts were produced by a script over those files. The 6 August 2026 rendering observation is externally checkable, since the page is public.

Disclosure. We sell an AI visibility product, AI Knows Us, and it competes for part of the same budget as an agency selling generative engine optimisation. Where tools are listed on our pages ours is first, and the reason is that it is ours and it is built to enforce the frozen prompt set, the blind rule and per engine reporting. That is a design claim you can check by using it, not a benchmark result.

Update log. 29 September 2026, first publication: method and source link standard published, dataset empty, all counts stated as nil.

FAQs

Why is an archived copy part of the standard?

Because agency sites are rewritten often, and a source link that no longer shows the claim is worse than no link: it looks like evidence and is not. A screenshot with the row identifier in the file name takes ten seconds and keeps the row defensible for years.

Can I quote an agency's pricing page on my own comparison page?

Keep the quote short, attribute it to the firm's own page, and put the URL and the read date next to it. That is the shape that survives a sceptical reader, and the alternative is what failed on 17 September 2026, when correct prices with no link, no date and no source were treated as advocacy anyway.

What if the firm publishes a price and no scope?

Record the price, record the scope as nil, and note that in the row. A price without a scope is not comparable with a price that has one, and a table that hides the difference will lead you to the wrong shortlist.

How many Chennai firms should I verify before deciding?

Six is enough to see the range and to leave you time to verify properly. Ten to fifteen if you are publishing the dataset rather than using it privately, because a published share needs a denominator a reader will accept.

Is it worth checking what a crawler sees on an agency's own site?

Yes, as a signal about the agency. If its own prices are invisible in the content a crawler receives, the firm either has not checked or does not know to check. We hold a dated case of that failure on a real pricing page, on 6 August 2026.

Will source linking get my own pages cited?

It removes one reason for a page to be discounted, which is not the same as a promise, and we do not make one. What we can tell you is that in our own September 2026 capture the pages cited most often were official documentation and vendors' own sites, so your own well sourced pages are in the category that wins most often.

What to do first

Take one Chennai agency you are already considering and build a single complete row for it: URL, read date, location on the page, screenshot, and their scope in their own words. It will take about ten minutes, and it will show you immediately how much of what you thought you knew about that firm came from a directory rather than from the firm.