A new model of marketing
Marketing that remembers. And gets stronger every month.
memarko is an agency that runs your company's marketing and website as one system: one memory of the business, sixteen areas of work, AI agents doing the routine and the parallel work, and a human approving every important decision.
The analysis shows what AI assistants and search engines say about your company on the questions your buyers ask: who gets named instead of you, and where the answers come from. It arrives by email.
The problem
The standard model of marketing is obsolete
It runs on people's memory, headcount and scattered tools: staff, contractors, chats, spreadsheets, decks and dozens of separate services. Everyone does their part. Nobody sees the whole business. This model took shape before today's technology and no longer competes with it. Here are its weak spots. They hold each other up.
- 01
Knowledge lives in heads, not in a system
The marketer knows the audience. The account manager remembers the agreements. The contractor holds the campaign history. One decision sits in a chat, another in a spreadsheet, the research in a deck. A person leaves, and part of the company's experience leaves with them.
- 02
Every new hand starts from zero, so you pay twice
A new employee, specialist or agency gets the same briefing again: who the customer is, what you sell, what was tried, what failed, which mistakes not to repeat. That onboarding was paid for once. You pay for it again at every change of hands.
- 03
The logic of decisions is lost, and guesses become rules
Months later the team remembers what was done, not why: which facts, what result was expected, what counted as success. Nobody wrote the threshold down before seeing the result, so any result can be explained afterward. 'Our audience doesn't care about this' lives on as law, and nobody recalls it was a guess.
- 04
Mistakes repeat, results don't accumulate
A rejected idea comes back under a new name. Old research gets done again. Campaigns run, content ships, reports land, tasks close, and what remains is numbers and files with no answer to what the company learned. Marketing produces activity, not intelligence.
- 05
Work runs in sequence, and time goes to waiting
One person researches and hands off. The next waits for approval. A third re-enters the task. Between them: meetings and retelling the context. Expensive specialists spend much of the day on searches, data pulls, spreadsheets and recurring reports. And one person cannot study a hundred competitors at once.
- 06
The work is graded by the person who did it
The agency reports on itself, and there is nothing to check the report against: no baseline was fixed. Inside the company it is the same. People defend their work instead of judging the result. Quality depends on how one person feels on one day.
- 07
Growth means more people, more meetings, more loss
More research, more materials, more channels means a bigger team. With it come costs, meetings, management time, lost information and longer approval chains. Headcount grows faster than the coherence of the work.
- 08
Everyone has AI, but it lives apart from the business
Most companies already have strong models. They are separate chats, where each person pastes their own version of the context. That AI knows nothing of the company's history, decisions or limits, and does not see what colleagues did. It gives a good answer to one question and never becomes part of the work.
The result
- You pay to move forward, but a large share of the money goes to returning to a point you already passed: re-onboarding, finding lost information, handing over context, manual routine, waiting between specialists, repeated research and mistakes made once before.
- Marketing works every month. It does not get stronger with every month.
It's not the people. The model is obsolete.
The system
Marketing that works as one system, not a set of services
A large part of marketing is memory, sequence, repeatability, control and discipline. In that, a machine is stronger than a person. Not cheaper. Stronger. We built an environment where marketing cannot forget its own decisions. People and AI agents work in it. The important decisions stay with a human.
- 01
All marketing memory in one system, not in heads
Your company gets its own working environment. In it is everything marketing has to remember: what you sell and to whom, how you differ, who the competitors are, what may be claimed in public, what was decided and why, what was done and what came of it. One fact in one place, with a date and a source.
- 02
A new hand, human or AI, starts from the accumulated context
Nobody gets the briefing again. Whoever joins reads the environment and starts from what the company already knows. A change of employee, contractor or even AI model resets nothing: the memory lives in the system, not in whoever works in it this month.
- 03
Decisions carry their reasons, and fact is kept apart from assumption
A choice that is expensive to undo gets its own record: the options, the facts they stood on, why this one, when to revisit. Everything else is a line in the journal. Every claim has a source; without one it is marked 'assumption'. The threshold of success is written before the measurement, not after.
- 04
A human can slip. The system checks before anything goes out
No marketer can hold a company's whole context in their head. An agent holds all of it, and a check stands before every outward action. A claim without a source is not published. A number without a measurement date does not enter a report. A rejected idea does not launch twice. A letter to a real person does not go without a yes. The mistake stops at the check, not at your customer.
- 05
Agents do the routine and the parallel work; people set goals and decide
Not one chat but a team of agents, dozens at once. One studies the audience, another the competitors, a third checks the site, a fourth drafts the texts, a fifth looks for errors in the others' conclusions. A hundred competitors means one per agent, and the results land in one system in one format. Scale grows by adding agents, not by hiring.
- 06
Baseline and threshold fixed before the work; the result is checked against a number
Before anything changes, a snapshot: where the company stands and how that was measured. Then a threshold: at what number we continue, at what number we stop. Then the actions. Then the same measurement, the same way. The report is checked against the criterion written before the work, and one agent's conclusions are verified by another. A zero stays in the record.
- 07
Every action leaves a trace, so the system gets stronger every month
Everything done becomes a journal line with a date and an outcome. A rejected idea stays rejected, with the reason and the condition for revisiting; it does not return in three months under a new name. In month six the system starts not where it started in month one, but with six months of accumulated knowledge.
- 08
AI works inside your context; new models plug in without a rebuild
An agent starts not from an empty chat but from the company's memory: goals, facts, bans, decisions, history. Models get stronger with every generation; the environment stays. A new model connects to the accumulated memory and continues from the same place. You do not rebuild your marketing at every step AI takes.
One business. One memory. Many hands. Human control.
Results
ShotBook: from zero to 726 users in 92 countries on a $0 budget
ShotBook is our own product, a catalog of video references for directors and editors. In May 2026 it had nothing: no audience, no ads, no history in search. The system ran the marketing; a human set the goals and approved the decisions. Each number below carries its measurement date.
- $0
- Marketing budget: a spending ceiling of zero, recorded August 31, 2026. No ads were bought; paid directories were declined.
- 40 → 45 → 96 → 372
- New users per month, May to August 2026. Closed months, read on September 1, 2026.
- 726
- Users from 92 countries between April 23 and September 8, 2026; 715 of them confirmed their email (98.5%).
- 34.2%
- Share of all sign-ups whose recorded source was ChatGPT: the #1 source, ahead of direct visits (30.9%) and Google (30.2%). Measured September 8, 2026.
- ×2.1
- Average month-over-month growth in new users, May to August 2026. Since July, at least double every month.
- 88%
- Of August's new users were active in the product within the month they joined: 328 of 372. Measured September 8, 2026.
- Numbers were read from the product's database and analytics on September 1 and September 8, 2026. Figures from September 2026 onward are not in our registry yet and are not shown.
- By the site funnel in August 2026, visitors arriving from ChatGPT signed up at 38.0%, from Google at 35.5%, from direct visits at 12.6%. Funnel data updated September 8, 2026.
- 'Source of sign-ups' is the channel recorded when the user arrived. It is not a claim that AI recommends the product.
- Your business is a different size in a different market. We do not promise you this curve. The order of work is the same: baseline, actions, the same measurement again, a decision by the number.
Five sweeps with zero citations
Every few weeks we put the same frozen set of questions to AI assistants and check whether ShotBook gets named: 46 cells, one per question and assistant. Sweeps on May 11, July 13, July 20, August 4 and August 11, 2026 returned zero. Before the sixth we wrote the threshold: 6 of 46 counts as a shift. The sixth sweep, closed September 7, 2026, returned 10 of 46, or 21.7%. The zeros stay in the report next to the growth. A frozen set of general questions and people arriving from ChatGPT with their own questions measure different things; the two numbers do not contradict each other.
Where the world is going
The world has moved to AI. Marketing still uses it as a chat
The buyer and the money have already moved. The way marketing is done has not. Each number below was checked against its primary source by a separate skeptic agent on September 7 and 8, 2026.
- 900M+
people use ChatGPT every week. A year earlier it was 400 million.
OpenAI, February 2026
- 2.5B+
people a month see Google's AI Overviews in search.
Google, May 2026
- 68%
of Google searches in the US end without a click.
SparkToro and Similarweb, June 2026
- 58%
of clicks lost by the #1 position when an AI overview is shown.
Ahrefs, February 2026
- +393%
AI traffic to US retail sites in Q1 2026 versus a year earlier. It buys 42% more often.
Adobe, April 2026
- 96%
of CMOs see marketing being rebuilt from the ground up. 8% use autonomous agents.
BCG, June 2026
- 16% → 36%
the share of marketing work done by AI, expected by 2028.
Gartner, May 2026
- 5%
of companies get substantial value from AI. 60% get almost none.
BCG, September 2025
Everyone has access to AI. The advantage goes to the company that rebuilt its system around it.
How it works
Week one builds the memory. Every week after makes it stronger
No publishing in week one, by design. Memory comes before action, and a baseline comes before any change.
Two tiers of thinking: the strongest available model plans, where decisions are made and mistakes are expensive; a cheaper model or a person executes a written brief. Your decision-maker decides. The system prepares, you approve.
- 1
You send the site. We run the analysis
Send your site address and your email. We check what AI assistants and search engines say about your company on the questions your buyers ask: who gets named instead of you, where the answers come from, whether crawlers can reach your site, how your listings and reviews look, which channels you have. The report comes to you by email.
- 2
A proposal built from what you have
Full marketing, or a package chosen by what your business has: a website, a product with sign-up, a person who can speak for the company. Scope and price come in a separate document after a conversation.
- 3
Week one: fifteen questions, a registry, a baseline, a plan
Fifteen questions about the business, one at a time, in your words. 'I don't know' is a normal answer: it is recorded with a date and becomes a task, not a hole. About 2.5 to 3.5 hours of your time spread over 7 to 10 days. Meanwhile we write the registry of what may be said about you in public, take the baseline of where you stand and how that was measured, and set the plan for the month. A week later, three questions about how the work should run: voice, form of contact, who decides what.
- 4
Sixteen areas, each wakes on a condition
Search, visibility in AI assistants, content, official channels, communities, email, partnerships and directories, ads and the rest are written into your environment from day one. They sleep. An area wakes when its condition is met, and the condition is a checkable fact: search needs public pages and console access; ads need a known customer value and a written spending ceiling. A channel your business does not have is marked not applicable and never pushed on you.
- 5
Preparation needs no permission. Anything outward waits for your yes
Drafts, plans and research are done without asking. Money, a message to a real person, a new public claim about you and anything irreversible wait for your yes. Silence is not consent: the task waits in your queue while the agents work on something else.
- 6
Threshold before the work, the same measurement after
Before anything changes we record where you stand and the number that counts as success. Then the actions. Then the same measurement, the same way. A zero stays in the report. Every action lands in the journal with a date and an outcome, and a rejected idea stays rejected with its reason.
The offer
Full marketing, three packages, two add-ons
Nothing switches on because you paid for it. Every area wakes on a condition of your business, and every package is built from areas that have ground under them in your company.
Full marketing
All sixteen areas of work in one environment with one memory. Each area wakes on its own checkable condition when your business has ground for it, and sleeps when it does not. You pay for a system that holds the whole field and works where there is something to work on.
- Sixteen areas written in from day one: search, AI visibility, content, official channels, communities, email, partnerships, directories, product, ads and the rest
- One memory: what you sell and to whom, what may be said in public, what was decided and why, what was done and what came of it
- Baseline before the start, threshold before the work, the same measurement after; a tree of metrics instead of one score
- Agents do the routine and the parallel work; your decision-maker approves anything that goes outward
If you need a part: packages by what you have
Search & AI
You have a website. There is no product with sign-up, or your own team already covers the rest.
- Search and visibility in AI assistants' answers: one job with two measurements, because a page built for search is also the source assistants quote
- Content written for the questions your buyers ask search engines and assistants
- Listings in directories and on review sites; partnerships
- Official channels, and support as a source of topics and reviews
Clicks from search by query group, and the share of questions where you are named, measured per assistant, each one separately.
Product
People buy your product without talking to a salesperson, and there is a sign-up.
- The path from sign-up to the moment of value and to payment
- Emails triggered by what the user does
- Invitations and referrals
- App store listings, and support as a source of topics
Share of users who reach the moment of value; actions after emails; invitations per user.
Voice
Someone in your company is ready to speak in its name.
- Communities where your buyers talk
- Press and public relations
- Talks, podcasts and live appearances
- Creators and authors; we prepare the plan, the texts, the answers and the letters, and your person speaks
Visits from communities; publications; actions after appearances.
Add-ons
Ads
A page of yours already brings actions without ads, and a spending ceiling with a switch-off condition is written down. Ads multiply what already works.
Outreach
You sell B2B, you have a list of recipients and a lawful basis to write to them. Messages go from a separate domain, and each batch goes out after your yes.
- A package is not sold to a business whose areas do not apply to it. No channel is promised that your business does not have.
- Ads are not sold on their own. They multiply a page that already works.
- Search is not sold without AI visibility. It is one job with two measurements.
- Adding an area later is a new agreement, not a move. It wakes in the same environment with all the memory already in place.
For Search & AI you pay only if your brand rises in AI answers above its starting step: per question, once for the rise and monthly to hold it. The other packages and the add-ons are billed monthly. Scope and price come in a document after the conversation. There are no prices on this site.
Straight talk
What we don't promise, and what you can hold us to
We don't promise
- That assistants and search engines will start naming you. The answer is formed by someone else's system under someone else's rules.
- Growth percentages. Without a baseline and a threshold, a forecast is fiction.
- Dates for results. A deadline for someone else's model is a guess dressed as a plan.
- Causality. If your numbers rise, we show what was done and when. We do not claim every rise was ours.
You can hold us to
- A baseline before the start: where you stand and how that was measured.
- A threshold written before the work, not after the result.
- A repeat measurement the same way, on the same frozen set.
- A journal line for every action, with a date and an outcome.
- A zero in the report instead of a nicer wording.
- A source on every claim; without one, the mark 'assumption'.
- Decisions stay with you. The system prepares, you approve.
Get the analysis
What do AI assistants and search say about your company?
Send your site address. We check the questions your buyers ask, see who gets named instead of you and where the answers come from, and reply by email with the analysis.
Questions
Questions we get
An agency that runs your company's marketing and website as one system. You get your own marketing environment: one memory of the business, sixteen areas of work written in from day one, AI agents doing the routine and the parallel work, and a human approving every important decision. Marketing that remembers what it decided and why.
Companies with a public website that sell on the world market, B2B or B2C. On your side, the owner or the marketing lead works with our lead. You provide the information, the access and the yes on public claims about you.
A regular agency keeps its knowledge in people's heads and grades its own work. Here the memory lives in your environment, not in whoever is on the team this month, and a new person or a new AI model starts from the accumulated context. The baseline and the threshold are written before the work, and the measurement repeats the same way, so the report is checked against a criterion instead of being written around the result.
A chat answers one question from whatever context you paste in. It does not know your decisions, your bans, your history or what your colleagues did yesterday. Our agents start from your company's memory, work in parallel by the dozen, check each other's conclusions, and stop before any outward action until a human says yes.
A check of what AI assistants and search engines say about your company on the questions your buyers ask: who gets named instead of you, where the answers come from, whether crawlers can reach your site, how your listings and reviews look, which channels you have. It is where the conversation starts, and it comes by email.
There are no prices on this site. Scope and price come in a document after the conversation. For Search & AI you pay only if your brand rises in AI answers above its starting step, per question: once for the rise and monthly to hold it. The other packages and the add-ons are billed monthly.
Once the scope and the price are agreed, week one begins: fifteen questions about the business, the registry of what may be said about you, the baseline, the plan for the month. No publishing in week one, by design. The date of the repeat measurement is written before the start. We do not name dates for results, because the answer is formed by systems we do not control.
AI agents do the routine and the parallel work inside your environment. Our lead runs the project and talks to you. Your decision-maker approves everything that goes outward. We work in the language of your market, and the set of AI assistants we measure follows the market you sell in.
See what the machines say about you
Send your site address. A person from memarko replies by email with the analysis.