Get recommended by AI

When buyers ask AI what to buy, Proofsource makes sure the answer is you, and proves which fix got you there.

chandan@proofsource.co proofsource.co

Proofsource AI Visibility screen for tesla.com

AI engines tracked10

On every number95% range

First full resultAbout 2 minutes

Traction

$10K MRR from about 15 customers, three weeks after launch.

MRR, 26 October

$10K

By week since launch on 7 October

Weeks 1–2: fill from Stripe

[$___]

$10K

Week 17–13 OctWeek 214–20 OctWeek 321–26 Oct

~15

Paying customers, most of them SaaS companies

$650

Average revenue per customer a month

131

Brands tracked across 80 accounts

Revenue: Stripe. Brands and accounts: PostHog. As of 26 October 2026.

The shift

Buyers now ask AI what to buy, and it names one or two companies.

51%

of B2B software buyers start research in an AI chatbot

69%

chose a different vendor because of one

G2 survey of 1,076 B2B software decision-makers, March 2026.

ChatGPT

Perplexity

Gemini

Claude

AI Overviews+5

[A real buyer question in a customer’s category]

Named in the answer

1[Competitor A]

2[Competitor B]

[Your company] isn’t named.The buyer moves on, and you never know.

Proofsource asks the same question on 10 AI engines.Example: add a real answer

The problem

Tools give a score and stop, and the score is noise.

A quarter of Perplexity’s recommendations change every day. A single score makes teams chase noise, and nobody can tell them whether a fix worked.

Day-over-day change in which brands get named, same question

Perplexity27%

ChatGPT18%

AI Overviews12%

Claude9%

Source: Proofsource sampling, as published on the pricing page.

The product

Track, understand, fix, prove.

Real Proofsource measurements of tesla.com from the team’s own account. The Prove example is illustrative.

AI Visibility: Tesla named in 48.3% of answers, ranked #1 of 9 — 01 · Track10 AI engines tracked, with ranges

Why you lost: 15 of 136 answers named a competitor and not Tesla — 02 · UnderstandWhy you lost each question

Fixes: questions ranked by impact, with drafts ready — 03 · FixRanked, drafted fixes

Verification of a shipped fix with re-checks on days 7, 14 and 28 (illustrative data) — 04 · ProveEach fix re-checked on days 7, 14 and 28

01 / 04

“A first full result takes about two minutes. Then we rank the gaps, draft the fixes, publish them, and re-check each one on days 7, 14 and 28.”

Presenting: → and ← step through the four screens before moving to the next slide; the arrows and the step list do the same by click.

On screen: the current dashboard (lcbasu/proofsource, run locally in demo mode) on tesla.com, the team’s own tracking account: AI Visibility, Why you lost and Fixes are real recorded measurements. The two fix drafts and the Prove panel’s shipped fix, its day 7/14/28 re-checks and the changed page are illustrative demo data, because no fix in that account has shipped yet. Say so if asked: “That verification screen is illustrative; we launched on 7 October, so our customers’ first day-28 results land in November.”

Facts: ten engines; a 95% range on every number; about 137 seconds for a full first scan (PostHog); 23,420 AI answers checked and 6,072 fixes found since launch.

Proof Fill before use: buyers counted, period, before and after, permission

One customer traced ₹20 lakh in sales to buyers who found it on ChatGPT.

Mention rate, with the range

Before

[X%]

→

After

[Y%]

Over [N] weeks, measured by Proofsource on the same questions.

Buyers who said they found it on ChatGPT

Before

[C]

→

After

[B]

About ₹20 lakh (about $[X]K) in sales traced to them.

In their words

[Named quote, if the customer permits]

[Customer logo][Name, role]

Method: the customer’s sales records and “how did you hear about us” answers, plus Proofsource visibility data.

Business model

Usage-based, about 5 cents per checked answer.

Start

$9 /mo

5 questions on ChatGPT, weekly

Land

$156 /mo

25 questions, 4 engines, daily

Grow

$630 /mo

100 questions

Average today ~$650

Expand

$2,929+ /mo

500 questions

On request

More engines and Managed

Per brand, per month. Customers grow by adding questions, engines and brands, with no plan change.

Go to market

SaaS companies buy directly; agencies are the channel we’re testing.

Paying customers today

~11 SaaS

3 consumer

1 agency

SaaS companies buy directlyConsumer brands1 agency running 3 client brands

Cold outreach funnelFill once tracked

Sent[___]

Replies[___]

Demos[___]

Trials[___]

Paid2as of 26 Oct

The agency offer

Sell AI visibility as your own service.

Their brand. Agencies resell Proofsource under their own name.
Their price. They set what their clients pay.
Their margin. They keep 80% of the markup.

Competition

250+ trackers; few do the work for non-experts and prove it worked.

Tracking and revenue attribution are becoming free features of bigger suites. Doing the work and proving which fix worked is not.

254 vendors in the category as of 26 October 2026. Placement reflects how each company describes itself.

↑ No expertise needed ↓ Needs experts ← Measures only Does the work → Measures only Does the work, no expert Does the work, for experts

Peec AItracking

Otterlytracking

S

SemrushSEO suite, now Adobe

ProofsourceSaaS and growing companies, with proof

T

Theolocal businesses

Profoundenterprise

A

AthenaHQmarketing teams

S

Sitefireenterprise brands

Why we win

What compounds: verified-fix data and accounts that grow.

Verified-fix data

[___]

Fixes shipped and verified since launch

Every fix re-checked after shipping teaches which fixes work, by engine and industry.

Accounts that grow

[___%]

Revenue growth within existing accounts since launch

Usage-based pricing grows with questions, engines and brands.

Work done for non-experts

[___%]

Customers shipping fixes without outside help

Measurement tools stop at a score; enterprise tools need an expert team.

“Features can be copied in weeks. Thousands of verified fixes, telling us which fixes work for which engine and industry, cannot.”

Fill before use: the three numbers, counted since launch on 7 October (fixes shipped and verified; revenue growth within existing accounts; share of customers who ship a fix without outside help). Three weeks of data is not a monthly rate, so say “since launch”.

Careful: we don’t have thousands of verified fixes yet. Say it as what we’re building: “Features can be copied in weeks. The data on which fixes work, by engine and industry, builds with every fix we verify.”

If asked: agency billing also compounds if the channel works, since clients, history and the agency’s income live in Proofsource (measure: agency clients billed through Proofsource checkout).

Vision

From AI visibility to the organic growth platform.

Proofsource today

Measure

Track what 10 AI engines say, with honest ranges.

Fix

Rank the gaps, draft the fixes, re-check each one.

Execute

Agents ship fixes end to end, on more site platforms and off-site.

Run it end to end

An autonomous system that runs search, social and AI for organic growth.

~$260M

a year from SaaS companies alone, at today’s $650 average (about 33,200 companies).

~$96B

search and SEO software budget in 2026, the pool AI visibility draws from (Precedence Research).

“AI visibility is the way in, because that’s where demand is most urgent. Every step uses the same data and the agents we already built at Grovio.”

Roadmap (18 months):

Measure and prove, now to January 2027: sign-in and first-scan fixes; four published case studies; agency workspace polish; revenue connected to analytics.
Execute, January to June 2027: agents ship fixes end to end on more site platforms (WordPress today; add Webflow, Shopify and others); off-site work made self-serve.
Search and social, July to December 2027: SEO and social content in the same loop; attribution to traffic, leads and revenue.
Organic growth platform, 2028: one system that plans, executes and proves organic growth across AI, search and social, built on Grovio’s agents.

Sizing: about 33,200 SaaS companies × $650 a month ≈ $259M a year (SaaS count from Fungies, an order of magnitude). SEO software $96.4B in 2026 (Precedence Research); definitions vary.

Team

Operators who did this by hand, and wrote every line of Proofsource.

CEO

Chandan Kumar

Growth at Mahindra Finance, IDFC FIRST Bank and ByteDance.

Co-founder, GTM and agencies

Amit Kumar

Uber, Slice; co-founded OG Club (acquired by Brinc).

The ask Fill before use: amount, use of funds, milestones

[Raising $X to reach Y]

A pre-seed usually funds 18 to 24 months.

Where it goesShare of the round

Fixes shipped for you

[2] engineers for execution and platform integrations

[__%]

SaaS outbound

[1] growth and sales hire

[__%]

Agency channel

[1] customer success and partnerships hire

[__%]

Product

Basith full-time on product and design

[__%]

What it buys, by [March 2027]

[$Z]MRR

[N]paying customers

[X%]of new customers from cold channels

chandan@proofsource.co

Appendix · Metrics

80 accounts in three weeks; weekly new accounts are falling.

New accounts per week since launch

44

24

12

Week 1Week 2Week 3

The launch spike faded, and we found sign-in failures and a weak first-scan step. [What we fixed and the result since]

23,420AI answers checked

245,717citations read

6,072fixes found

~137 sfor a full first scan

37%of trials finish a first scan (activation definition still to settle)

Too earlyfor retention: we launched on 7 October

1,599visitors from the US, India and about a dozen other countries

12.8%click sign-up in India, our best-converting market

PostHog, last 30 days, as of 26 October 2026.

Appendix · Pricing maths

Price per checked answer, tier by tier.

Tier	What it buys	Checked answers a month	Price per answer	Our cost per answer
$9	5 questions on ChatGPT, weekly	[___]	[___]	[___]
$156	25 questions, 4 engines, daily	about 3,000 (25 × 4 × 30 days)	~5.2¢	[___]
$630	100 questions	[engines and cadence]	[___]	[___]
$2,929+	500 questions	[engines and cadence]	[___]	[___]

AI and data costs today

About $7 to $43 a day in total, across all accounts.

Gross margin per account

[___%]

Appendix · The market

254 vendors, and the money is concentrating.

Vendors by segmentCompanies

Long-tail trackers (directory-listed)118

Mid-market and small-business specialists (ours)29

SEO suites adding an AI module25

Agencies and services23

Venture-backed platforms18

Content and SEO tools adding GEO10

Commerce and vertical specialists9

Enterprise marketing suites8

Analytics and infrastructure layers8

Free checkers6

223 active, 7 acquired, 12 possibly defunct, 12 unverified. Only 73 disclose funding.

Rounds and acquisitions

Profound	$180M Series D at $1.8B	Sep 2026
Bluefish	$43M Series B	Apr 2026
Brandlight	$30M Series A	2026
Peec AI	$21M Series A	Nov 2025
Evertune	$15M Series A	Aug 2025
Searchable	$14M at about $85M	May 2026
geoSurge	$12M seed	Jul 2026
Semrush	Acquired by Adobe, about $1.9B	Apr 2026
Scrunch	Acquired by Sitecore	Jun 2026
Hall	Acquired by Tracksuit	Jul 2026
Meridian	Acquired by Searchable	Sep 2026

Vendor roster compiled 26 October 2026 from vendor sites, press and directories.

Appendix · Market size

SaaS alone is about $260M a year at today’s price.

Bottom-up: what Proofsource can reach

Segment	Count	Price assumption	A year if all bought
SaaS companies	~33,200 (~17,000 in the US)	$650 a month, today’s average	~$259M
US agencies reselling	100,202 (IBISWorld)	5 client brands at $156 a month	~$938M

Realistic shares: 1% of SaaS companies is about 330 customers and $2.6M a year; 5% is about 1,660 customers and $13M. 1% of US agencies with five brands each adds about $9.4M.

Search and SEO software budget

$96.4B

in 2026, about $295B by 2035 (Precedence Research). The pool AI visibility draws from.

GEO specifically

$1.09B

in 2026, $17.15B by 2034 (Dimension Market Research, via Superlines).

Typical tool spend today

$337 /mo

across 30+ tools (Rankability, via Superlines). Our $650 average sits above it.

Appendix · YC’s companies in this space

Twelve YC companies, and how we differ.

Company	Niche	Proof	How we differ
AthenaHQ (W25)	Marketing teams, startups to Fortune 500	Share of voice 2.0% to 12.6% in 60 days	[one line]
Sitefire (W26)	Enterprise brands, Europe	BMW, DWS; Jerry +78% AI referral traffic	They sell tracking plus content agents to enterprises; we sell to SaaS and growing companies, with proof of which fix worked
Theo (F26)	Small local businesses, website plus visibility	10,000+ sites; $29.99 to $69.99 a month	They sell websites to local businesses; we sell to SaaS and growing companies
Wildcard (W25)	E-commerce and AI shopping	20+ brands; presence 1.16% to 23.25%	[one line]
Unusual (F24)	Brands and scale-ups (“AI Relations”)	Change.org, Astronomer	[one line]
The Prompting Co. (S25)	Dev tools and software used by agents	NVIDIA, Google, Cloudflare logos	[one line]
Manicule (X26)	Dev tools (“DevRel for agents”)	$53K a month, 100% monthly growth	[one line]
Scope (X26)	Software used by agents	A decacorn design partner	[one line]
Armature (X26)	Dev tools picked by coding agents	17,000 agent runs measured	[one line]
Rainbrand (X26)	Brand context for agents	175,000+ assets generated	[one line]
Openbenchmarks (F24)	Benchmarks for vendor choice	Accepted as OpenFunnel	[one line]
OpenSEO (F26)	Open-source SEO and AI visibility	3,000+ users	[one line]

Appendix · More case studies

Three more case studies, one slide each.

Case study 2[SaaS customer: before and after, method, permission]

Case study 3[SaaS customer: before and after, method, permission]

Case study 4[Customer: before and after, method, permission]

Appendix · Questions to rehearse

Hard questions, short answers.

Question	Short answer
How much of the $10K is from your network?	About 13 of 15 customers were design partners from our network who chose to pay; 2 came cold. We’re now proving cold acquisition.
Why are weekly sign-ups falling?	The launch spike faded, and we found sign-in failures and a weak first-scan step. [What you fixed and the result since.]
How do you prove a fix worked when answers change daily?	The same questions measured before and after, with 95% ranges, judged on day 28.
Why won’t Profound or Peec copy you?	They can copy features. They can’t copy our data on which fixes worked, or accounts that already grow with us.
Won’t Google or OpenAI give this away?	They’ll show companies their own data. They won’t fix it, prove it, or work across rival engines.
Why you?	We did this work by hand for a decade, and wrote all the code ourselves.