Foundations
Am I too early for AEO?
By Arnav Mukherjee, founder of TofuBofu · July 31, 2026
TL;DR
If you are hunting your first ten customers, yes, you are early, and I would rather tell you that than sell you a subscription. AEO is demand capture and it needs a track record to work on. Around $100K ARR is the natural entry point, because that is when you first have both the customers who generate proof and the positioning an engine can match. The mechanism genuinely favours incumbents: in our data the top 10 named companies took 42% of all mentions while 64% of companies appeared exactly once. Smaller firms win by picking narrower questions, not by fighting for the broad ones.
I run two companies, and the second one taught me this the hard way. When we first pointed our own scanner at TofuBofu, it returned zero across every buying question. Not low. Zero. And the honest reason was not that our content was bad. It was that we were three weeks old. There was nothing about us anywhere for an engine to retrieve, so no amount of optimisation was going to produce a citation.
That is the question underneath "am I too early." Founders ask it in two versions: "I want my first ten B2B customers, is this for me?" and "I am at $100K ARR heading to a million, is now the moment?" Those have genuinely different answers, and the honest one for the first version costs us a sale. Here is the reasoning, and the data that shapes it.
What AEO actually needs to work at all
Strip away the vocabulary and AI recommendation runs on three inputs. Getting the stage question right is mostly a matter of asking which of the three you currently have.
Something to retrieve. An engine can only recommend a company it can find information about. That means content, but more importantly it means content that exists somewhere other than your own website. A brand-new company is not invisible because it is bad. It is invisible because it is genuinely absent from the corpus.
A reason to trust it. This is the corroboration layer, and it is the one that cannot be shortcut. Review sites, community discussion, independent mentions. The measured effect of breadth here is large, with Profound putting brands on four or more platforms at 2.8x the citation likelihood. Now count how many of those platforms you can populate without customers. Reviews need buyers. Community threads need users. Case studies need results. A pre-revenue company is not behind on this work, it is structurally locked out of it.
A category to be matched against. When a buyer asks for "the best X for Y", the engine has to decide you are an X. If your own positioning is still moving, which it should be early on, there is no stable thing to match. This is the input founders most consistently underrate, and the cheapest one to fix.
A pre-revenue company typically has none of the three. A company at $100K ARR usually has the third and can build the second. That gap is the whole answer.
The data: this is a winner-take-most channel
Across 75 completed scans covering 44 domains, we logged every company an engine named in response to a buying question. That produced 7,333 individual mentions across 1,329 distinct firms. The shape of that distribution decides how realistic your ambitions should be.
| Finding | Value | What it means for you |
|---|---|---|
| Top 10 firms' share of all mentions | 42% | The head is small and crowded with incumbents |
| Firms named exactly once | 64% | Appearing once is common and means little |
| Scans returning zero mentions | 31% | Starting from nothing is the normal case |
| Median presence across buying questions | 8% | Even visible firms are absent most of the time |
TofuBofu first-party data. 75 completed scans across 44 domains, June 24 to July 27, 2026. Self-selected sample: companies that run a visibility scan usually suspect they have a problem, so treat the zero rate as a floor rather than an industry average.
Read those four rows as one statement: AI recommendation is a short list, and short lists concentrate. Ten blue links had room for ten companies. An AI answer names four or five, and it prefers the ones it has seen corroborated in many places, which is a description of an incumbent.
I want to be careful about what this data cannot tell you, because the honest limits matter. We do not hold reliable revenue or headcount figures for the companies we scan, so I cannot show you a clean chart of visibility against company size, and I am not going to imply one from industry labels that customers typed in themselves. What the concentration figure does establish is that the mechanism rewards accumulated proof, and accumulated proof correlates with age and scale. That is an argument about incumbency, not a measurement of it.
Find out where you sit on that curve
A free scan shows whether six AI engines name you on your real buying questions, and exactly which companies are taking the mentions instead.
Run your free scanThe answer by stage
Four stages, and the honest recommendation at each. Note that two of them say do not buy anything.
Pre-revenue, pre-launch
Not yet. Do the free half.
You have nothing for an engine to retrieve and no proof for it to trust, so any measurement will read zero and stay there. But the structural work is nearly free and it is much cheaper now than retrofitted later: state your category in plain language on your homepage, add basic Organization and Service schema, make sure robots.txt is not blocking AI crawlers, and write real answers to the questions your first prospects ask. Skip the subscription entirely.
Hunting your first ten customers
No. Go sell directly.
AEO is demand capture: it catches buyers who are already looking and do not know you. At this stage you need demand creation, which means founder-led outreach, your network, and being in the rooms your buyers are in. Those are faster, more controllable and give you the customer conversations that later become your content. The one exception worth making is that every early customer should be asked for a review on whichever platform your category actually uses, because that is the corroboration layer starting to build.
Around $100K ARR, scaling up
Yes. This is the entry point.
You now have the two things you lacked: customers who can leave real reviews, and a category position stable enough for an engine to match. You also still have the problem, which is that strangers cannot find you. Start narrow. Pick the ten buying questions your best customers actually asked before they bought, answer them properly, get listed and reviewed where your category is reviewed, and measure whether engines start naming you. This is the stage where the compounding starts, and starting here rather than at $1M is a real advantage.
$1M ARR and beyond
Yes, and the target changes.
At this point you are probably already named occasionally, so the job shifts from existence to position: which questions do you win, which do competitors win outright, and where does an engine frame you against the wrong set of rivals. This is also the stage where the head of the distribution becomes reachable, because you have enough corroboration to compete on broader category questions rather than only long-tail ones.
Which companies actually do well
The most common version of the stage question is really a size question: is this for big companies or small ones? Having looked at a lot of these scans, size is not the variable that predicts results. Three other things do, and a small company can have all three while a large one lacks them.
A category you can finish in one sentence. The firms that do well can complete "we are the X for Y" without hedging. The ones that struggle describe themselves as full-service, end-to-end or a trusted partner, which gives an engine nothing to match a specific query against. This is the single most common problem I see, and it is a positioning problem wearing a technology costume.
Proof that lives somewhere you do not control. Reviews, community mentions, coverage, directory presence in the places your category is genuinely evaluated. Anything on your own domain is a claim; anything off it is evidence, and engines weight evidence.
Buyers who research before they contact you. If your market's buying process is dominated by procurement, tenders or personal introductions, the AI answer sits outside the purchase path and results will be thin regardless of how well you execute. If your buyers open a chat window and ask who they should be talking to, the answer sits directly in the path.
Which brings me back to the founder chasing their first ten customers. You are not too early to care. You are too early to buy. Get your category language sharp, make the site machine-readable, and ask every one of those first ten for a review. Then, at the point where strangers start needing to find you, the groundwork is already done and the measurement finally has something to measure.
Frequently asked questions
Am I too early for AEO if I am chasing my first ten customers?
For getting those specific ten customers, yes, almost certainly. AEO is a demand-capture channel and it works by being present when a stranger asks an engine for a recommendation, which requires a track record you do not have yet. Your first ten customers come from your network, direct outreach and founder-led selling, all of which are faster and more controllable. What is worth doing at this stage is the structural half: make sure your site is machine-readable and your category is unmistakable, because that work is cheap now and compounds later. Do not buy a monitoring subscription to watch a number that will read zero for months.
Is 100K ARR the right time to start AEO?
It is close to the ideal entry point. At 100K ARR you usually have the two things AEO needs and a pre-revenue company lacks: real customers who can leave real reviews, and a proven category position you can state clearly. You also still have the problem AEO solves, which is that buyers who have never heard of you cannot find you. The move at this stage is not a big content programme, it is corroboration: get on the review sites your category actually uses, publish the specific buying-intent answers your sales calls already contain, and measure whether engines start naming you.
Does AEO work better for big companies or small ones?
The mechanism rewards incumbency, which favours the established. In our first-party scan data, engines named 1,329 distinct companies across 7,333 mentions, but the top 10 companies took 42 percent of all mentions while 64 percent of named companies appeared exactly once. That is a steep power law and the head of it is occupied by large, well-corroborated brands. The opportunity for smaller firms is not to beat those brands on broad category questions, it is to own narrower questions where specificity beats scale and where the large incumbents are too generic to be the right answer.
What kind of company gets the best results from AEO?
Companies with a sharp, nameable category position, real third-party proof, and buyers who research before they buy. Those three together matter far more than size or sector. A firm that can finish the sentence "we are the X for Y" gives an engine something to retrieve and match. A firm with reviews and independent mentions gives the engine a reason to trust it. And a market where buyers compare vendors before contacting them is a market where the AI answer sits directly in the purchase path. Where any one of those three is missing, results are much slower.
Should a pre-revenue startup spend money on AI visibility tools?
Generally no, and this is against our own commercial interest to say. Pre-revenue you have no reviews, no track record and often no settled positioning, which means there is very little for an engine to retrieve and your score will sit near zero regardless of what you pay to watch it. Do the free structural work instead: clear category language on the homepage, basic schema, an unblocked robots.txt, and a handful of genuinely useful answers to buying questions. Start measuring when you have customers who could plausibly be cited.
How long before AEO is worth the investment at an early stage?
Think in terms of prerequisites rather than months. The investment starts paying when three things are true: you have enough customers to generate third-party proof, you have a category position specific enough that an engine can match a query to it, and you are getting buyers who found you rather than were introduced to you. Companies that hit those conditions see movement in weeks on search-grounded surfaces. Companies that start before them can spend for a long time with nothing to show, because the constraint is not effort, it is the absence of anything for an engine to cite.
Is it too late to start AEO if my competitors already dominate AI answers?
No, but you should change the target. Attacking a broad category question that an established competitor already owns is expensive and slow. The realistic path is narrower questions: a specific use case, a specific buyer size, a specific integration, a specific geography. In our data the most-named companies concentrate on broad category questions, which means specific long-tail questions are frequently uncontested. Winning ten narrow questions your best-fit buyers actually ask is worth more commercially than losing one broad question to a company ten times your size.
Sources and further reading
- TofuBofu first-party scan data: 7,333 company mentions across 1,329 distinct firms, from 75 completed scans covering 44 domains, June 24 to July 27, 2026. Self-selected sample skewed to B2B services and SaaS
- Profound research: brands present on four or more platforms are 2.8x more likely to be cited, which is why corroboration is the stage-gating input
- G2 2026 AI Search Insight Report: 51% of B2B buyers start research on an AI chatbot, and one in three chose a vendor they had not heard of before AI recommended it
- Kevin Indig and G2 on review volume and AI citations: 10% more reviews associates with roughly 2% more AI citations, the mechanism behind the corroboration advice