Pune marketing agencies compared: services, public pricing and selection method

The count that gives a city directory its value is the number of named agencies whose published service and location details were checked against their own sites, with the cutoff date attached. Ours is 0, as of 29 September 2026. We have checked no Pune agency, so this page names none and compares none. What it publishes is the verification method in full: the eight checks that turn a claim into a recorded fact, the rule for what counts as primary evidence, and the price recording standard. The reason we hold the line on primary evidence is a measured one, from another sector: on 17 September 2026 Claude read two comparison pages in legal technology whose competitor prices were correct and carried no link, no date and no source. It checked them independently, found them correct, and still treated the pages as advocacy and used official sources instead.

Most city directories fail at the same point. They take a service list and a location from a third party listing, put it in a table, and present it as verified. The firm may well have moved, dropped a service, or never offered it. Checking against the firm's own site is slow, and it is the only thing that makes a directory a source rather than a copy.

Scope and definitions

• Pune means a publicly stated office address inside Pune Municipal Corporation or Pimpri Chinchwad Municipal Corporation limits, recorded separately so a reader can see which. Areas where agency offices are commonly found include Baner, Kharadi, Viman Nagar and the Hinjawadi information technology park.
• An agency means a firm that publicly states it sells marketing services to clients. An in house team, a reseller with no service page, and a single freelancer with no company page are all out of scope.
• A service is counted only where the firm's own site names it. Search optimisation, paid media, content, social media, and generative engine optimisation are counted separately, never bundled into digital marketing.
• Checked means checked against the firm's own site, with a URL and a read date recorded per claim. A directory profile is not the firm's own site.
• Cutoff date means the latest date on which any row was verified, and it is printed with every count.

Method and sources

What counts as primary evidence, in order of strength. Record which level each field came from, because a reader who knows the level can decide how much to trust the row.

• Level one. The firm's own website page, read on a recorded date, with a screenshot saved.
• Level two. A public filing, such as the Ministry of Corporate Affairs register for the company name, registered office and directors.
• Level three. A document the firm published itself, such as a rate card or a scope PDF, with the caveat that these are often undated.
• Level four. A third party listing, such as a directory or marketplace. Usable to find a candidate, not to verify a claim.
• Level five. Something somebody told you. Not evidence, and not recorded as a fact at any point.

The eight checks. Run all eight per firm. Each one produces a pass, a fail or a nil, plus the URL and the date.

• One. Does the site state a full postal address, not only a city? Record the address and which municipal limit it falls in.
• Two. Does the address on the site match the registered office in the public filing? Record both. A mismatch is not misconduct, it is a fact to note.
• Three. Is every service in your table named on the firm's own service pages? Delete any service you cannot find there, however many directories list it.
• Four. Does each service page state a scope, naming what is included and excluded? Record the presence or absence, and quote nothing without a link.
• Five. Is there any evidence the site has been updated in the last twelve months? A dated post, a copyright year, a changed page. A site frozen for three years is a finding about the firm.
• Six. Is there a named human with a role and a working contact? Test the contact and record the date you tested it.
• Seven. Is any claim on the site checkable, and does it check out? This is the check that protects you most. On 6 August 2026, in legal technology, a headline figure on a company's main site could not be true, and when Claude checked it against public numbers it advised a buyer against the product, after which the company's accurate claims stopped counting for that answer. An agency whose headline numbers do not survive a check is a risk you can see from outside.
• Eight. Does the content a crawler receives contain the claims you just read? Fetch the page without running scripts. On 6 August 2026 a pricing page rendered its prices only after scripts ran, so what a crawler received contained no prices at all, and the prices assistants quoted came from an app store listing instead of from the company's own site. If an agency cannot get this right on its own site, ask how it will get it right on yours.

Named agency profiles

Agencies named and verified against their own sites: 0. Cutoff date: 29 September 2026. This section is empty, and it is dated, which is the honest version of not having done the work.

Here is the nearest dated evidence we hold, with sectors named so that nothing is read across. None of it is a measurement about Pune, about agencies, or about marketing services.

• 0 of 6 questions in which 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.
• The most cited domains were the assistants' own documentation and the vendors' own sites. AI visibility tooling, aiknowsus.com, September 2026, Perplexity, counted across 24 captures. This is the measured reason the eight checks all point at the firm's own site.
• Published coverage lists beat a better product. Legal technology, clawlaw.in, 17 September 2026, Claude. Two enterprise vendors were ranked above a smaller one specifically because they publish explicit court and tribunal coverage lists, and the assistant noted that its ordering reflected price transparency and source authority rather than product quality.
• Our own count on our own category questions was 0 of 6, AI visibility tooling, aiknowsus.com, September 2026, Perplexity, with 161 recorded statements across 24 batches and 72 conversations that we were not cited. We publish our own zero because a page that asks you to collect evidence should show its own.

Comparison by business need

We rank nothing, because we have scored nothing. Six needs, and for each one the checks and questions that decide it.

• A manufacturing or engineering firm selling to industrial buyers. Weight checks three, four and seven. You need a firm that can write a scope for a long, technical sales cycle and whose own claims survive a check.
• A software company with an international customer base. Ask which countries and languages any measurement covers, and require per country reporting. An answer measured in India is not an answer measured elsewhere.
• An education or training business with seasonal demand. Ask about minimum term and the ability to pause. Get it in writing with a date, because a verbal assurance about pausing is worth nothing in month seven.
• A local services business with one location. Weight checks one, two, five and six. A firm that keeps its own address, its own filing and its own site consistent is the firm most likely to keep your listings consistent.
• Anyone buying generative engine optimisation. Ask for the frozen prompt set, check your brand appears in none of it, and ask for the run count arithmetic. The reason is measured: on 27 July 2026 two runs of the same 78 questions happened on the same day for clawlaw.in, and the run whose wrapper named the brand came back ranking it first on almost every question, while the blind run put the company second by breadth and absent from the questions it most wanted to win. The first run was discarded.
• Anyone who has been burned once already. Put raw data delivery into the contract, monthly, in your own storage, and make the first month's deliverable a dated baseline with numerators and denominators rather than a strategy deck.

Price data and its limitations

No agency prices appear on this page, on 29 September 2026, because we have verified none. We do not restate prices we have not read on a date, and we do not restate prices we have read without the date and URL beside them. Record yours in these eight fields.

• Service, in the firm's own words.
• Price, with currency.
• Shape: fixed retainer, starting from, band, hourly, per project, percentage of media spend, or packaged tier.
• Evidence level, one to five from the list above.
• Scope stated with the price.
• Minimum term and notice period.
• URL and read date.
• Screenshot file name.

Three limitations of any price dataset like this, stated before you draw a conclusion from yours.

• A published price is a starting position, not the price you will pay. Scope negotiation moves it, and the dataset cannot see that.
• Firms that publish prices are not a random sample of firms. Publishing is itself a business decision, so a range built only from publishers describes publishers.
• A price with no date is already wrong. Record the read date per row and recheck quarterly, keeping the old rows.

Dataset, corrections, and update history

Dataset. Firms checked: 0. Firms with services verified on their own site: 0. Firms with a public price: 0. Cutoff date: 29 September 2026. When rows exist, each will carry evidence level, URL, read date and screenshot, and the whole sheet will be published rather than summarised.

Sources for every figure quoted 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 observations about a page's rendering and an unsupportable headline figure are externally checkable, because the pages involved are public.

Disclosure and conflict of interest. We sell an AI visibility product, AI Knows Us, which competes for part of the same budget as an agency selling generative engine optimisation. Where tools are named on our pages ours is listed first, and the one line 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, and we have never tested ourselves against an agency's service.

Corrections and updates. Write through the contact page if you run a Pune agency and want to be verified under this method, or if something here is wrong. Corrections are logged with a date and the previous value. Version: 29 September 2026, first publication, empty dataset, method published.

FAQs

Why not just use a directory list and save the work?

Because a directory tells you what the directory says. In the one dated case we hold, across six questions in legal technology on 17 September 2026, no third party review or directory source reached any answer at all, and a well known review site was discarded because its list was of American products. Your own verification is the only evidence you will be able to defend.

How long do the eight checks take per firm?

Time your first firm and use that as your estimate rather than trusting anybody's figure, including ours. What we can tell you is that check three, matching services to the firm's own service pages, removes more wrong rows than the other seven together.

What if the firm's site and its public filing disagree on the address?

Record both, flag it, and ask. Companies move and filings lag. The point is not to catch anybody out, it is to make sure your table does not silently assert one address as verified when your evidence shows two.

Is Pimpri Chinchwad the same market as Pune city?

Treat them as two columns and publish both counts. Whether they are the same market for your purposes is your decision, and it is a decision you can only make if the data lets you see the split.

Does any of this help me appear in AI answers?

Indirectly, and here is the honest version. The domains cited most often across our September 2026 capture were official documentation and vendors' own sites, so what your own site publishes matters. That is not a promise of a position in an answer, and we do not make one.

What is the fastest signal that an agency will report honestly?

Ask for last month's report for another client with the client details removed. If it contains numerators, denominators, engine names and dates, you are dealing with somebody who measures. If it contains a score with no denominator, you have your answer in five minutes.

What to do first

Pick three Pune agencies and run only check three and check seven on each: is every service you expect named on their own service pages, and does their most prominent claim survive a check against a public source. Those two checks take under an hour for three firms and they remove the candidates that would have cost you a quarter.