B2B software buyers start in an AI chatbot.
Half of respondents to G2’s August 2025 survey said they start their software buying journey in an AI chatbot instead of Google Search.
G2 · 6 Oct 2025 ↗Answer engine optimization (AEO)
We get tech startups recommended by AI.
Twozeds runs AEO for seed and Series A tech companies. We get you named by ChatGPT, Gemini and Perplexity, and picked by personal agents like Instinct, Muse, Poke and Grok.
Get discovered on
100 buyer questions per engagement4 answer enginesFree check back in 72 hours
The shortlist is written before buyers reach your site.
Buyers ask an AI assistant, get two or three names, and start from there. If you are not one of the names, you are not in the deal.
Fewer clicks for the top Google result under an AI Overview.
Ahrefs found AI Overviews correlated with a 58% lower average desktop click-through rate for the top-ranking page, relative to its estimated no-Overview baseline.
Ahrefs · 4 Feb 2026 ↗Of searches with an AI summary end in a click on a normal result.
In Pew’s study, visits to Google results with an AI summary led to a traditional-result click 8% of the time, versus 15% without a summary.
Pew Research Center · 22 Jul 2025 ↗SEO gets you ranked.
AEO gets you named.
You rank on Google. AI names someone else.
Rankings and recommendations are two different lists. Being first on one does not put you on the other.
You checked ChatGPT once. That is not a measurement.
Answers change with the wording, the engine and the day. One screenshot tells you very little.
Your site reads well to people. AI gets half of it.
Many AI crawlers do not run JavaScript. If your pages need it to load, they arrive empty.
Named by AI.
Picked by agents.
An AI answer tells a buyer who to consider. A personal agent goes further: it researches, chooses and buys for them.
What are the best customer support platforms for an early-stage SaaS company?
Depending on your team’s needs, you could compare:
- ExampleCo AlphaFor a small support team.
- ExampleCo BetaFor a shared support inbox.
- ExampleCo GammaFor product-led support.
Your company: not mentioned
AI answers
A person asks, reads the answer and decides.
The goal: be named.
How we measure it: 100 buyer questions, counted by how often you appear.
Find me a support tool for a five-person team and start a trial.
Your agent picked
ExampleCo Alpha
- Price confirmed
- Free trial confirmed
- Sign-up completed
Your company: pricing not confirmed
Personal agents
The agent finds options, picks one and completes the order or sign-up.
The goal: be picked, and be possible to finish.
How we measure it: 10 real tasks, each scored as considered, chosen and completed.
Fictional companies. Not live results.
Where an agent drops you
Not found.
It never comes across you.
Not chosen.
It can't confirm your price, delivery or terms.
Not completable.
It gets stuck at sign-up or checkout and moves to the next option.
What we do, in order.
Five stages. Each one ends with something you can open and read.
Find the questions your buyers ask.
We collect the way buyers phrase things in your sales calls, reviews, Reddit threads and competitor comparison pages, and turn it into a fixed set of 100 questions.
Sales callsReviewsReddit threadsComparison pagesQuestion set
YOU GET
the full question set, yours to keep.
Measure what AI says about you today.
We run every question 3 times on ChatGPT, Gemini, Perplexity and Google AI Mode, because a single answer is noise. Then we read the answers. We also give Instinct, Muse, Poke and Grok 10 real tasks and record whether it considers you, chooses you and can finish.
Times namedWho is named insteadAccuracy checkSources citedAgent tasks
YOU GET
a baseline report with every question and every answer.
Make your site easy for AI to read and quote.
We fix what stops AI reading your pages, then write or rewrite the pages your buyers' questions point at.
Crawler accessLoads without JavaScriptStructured datallms.txtProduct, comparison and pricing pagesFAQSign-up and checkout an agent can finish
YOU GET
the changes shipped, or a ticket list for your developer.
Get mentioned where AI looks.
AI leans on what other sites say about you more than what you say about yourself. We map the sources it cites in your category and work to get you into them.
Source mapComparison articlesReview sitesCommunitiesYouTube and podcastsPress
YOU GET
a source map and a monthly list of mentions earned.
Re-test, report, repeat.
Same questions, same engines, every month. We report what moved, what did not, and what we do next.
Monthly readoutPer-engine resultsCompetitor movementNext month's plan
YOU GET
a monthly readout and next month's plan.
Diagrams use fictional data.
What you'll have in your hands.
Three documents. Written by us, in plain language, with a recommendation in each.
AI names you in 27 of 100 buyer questions. It names ExampleCo Alpha in 50.
The three questions costing you most
| Question | Named instead |
|---|---|
| Which support platform fits a small team? | ExampleCo Alpha |
| What replaces a shared inbox? | ExampleCo Beta |
| Best support tool for a product-led company? | ExampleCo Alpha |
What AI gets wrong about you
Quotes pricing from your old plans.
Says you have no free plan.
What we recommend
Fix the pricing page first, then publish a comparison page against ExampleCo Alpha. Those two changes touch 31 of the 100 questions.
Sample documents. Fictional company and numbers.
Visibility isn’t just a technical problem.
It’s a distribution problem.
One team for your site and your mentions.
AI pulls from articles, review sites, YouTube, Reddit and press. Twozeds is a distribution studio, so the team that fixes your site also makes and places the content that gets you talked about.
The method is on this page.
Question counts, engines, runs, cadence. You see every question we ask and every answer we get.
No guarantees we can’t keep.
Nobody controls what a model says. We control the inputs and show you the numbers every month.
How this compares
| Twozeds | A tracking tool | An SEO agency | |
|---|---|---|---|
| Who does the work | We do | You do | They do |
| What gets measured | Whether AI names you, question by question | Whether AI names you | Rankings and traffic |
| Site fixes | Shipped by us or handed over as tickets | Not included | Usually included |
| Mentions on other sites | Made and placed by the same team | Not included | Mostly link building |
| What you receive | Every question and answer, plus next month's plan | A dashboard | A rankings report |
Twozeds
- Who does the work
- We do
- What gets measured
- Whether AI names you, question by question
- Site fixes
- Shipped by us or handed over as tickets
- Mentions on other sites
- Made and placed by the same team
- What you receive
- Every question and answer, plus next month's plan
A tracking tool
- Who does the work
- You do
- What gets measured
- Whether AI names you
- Site fixes
- Not included
- Mentions on other sites
- Not included
- What you receive
- A dashboard
An SEO agency
- Who does the work
- They do
- What gets measured
- Rankings and traffic
- Site fixes
- Usually included
- Mentions on other sites
- Mostly link building
- What you receive
- A rankings report
Common questions.
Something else? work@twozeds.com
Is this just SEO with a new name?
Some of it starts in the same place: AI can’t recommend a site it can’t read. But SEO is measured in rankings and clicks. AEO is measured in whether you are named in the answer, and much of the work that moves that happens off your site.
Can you guarantee ChatGPT will recommend us?
No, and nobody can. We control what AI can find and read about you, and we measure the result every month.
How do you measure it?
We ask the same 100 buyer questions on 4 engines, 3 times each, every month, and count how often you are named. You get the full list of questions and answers.
How long does it take?
Engines that search the live web can pick up changes within weeks. What a model has memorised changes more slowly. Plan for three months before judging it.
Can I see results before I pay?
Yes. The free check is run on your company, so the first results you see are your own.
What do you need from us?
One 45-minute call, access to your site or CMS, and read access to your analytics.
Do you cover personal agents?
Yes. We test Instinct, Muse, Poke and Grok with real tasks alongside the AI answer engines. Agents read your own pages for price, stock and terms, then try to complete the order or sign-up, so we check all three. We add agents to testing as we get access to them.
Find out if AI recommends your startup.
We ask 25 of your buyers’ questions on ChatGPT, Gemini, Perplexity and Google AI Mode and send you the results within 72 hours.
- How often you are named
- Who is named instead, question by question
- What AI gets wrong about you
- The first fix we would make.
Free. The results are yours whether you hire us or not.
Or book a call ↗