Claim Engine · Generative Engine Optimization
Change what AI says about your category.
When buyers ask ChatGPT or Google's AI which vendor to pick, the answer is built from claims: who's easiest to set up, who gets expensive, who fits a team like theirs. We map those claims, trace where they come from, and help you own the ones that decide the shortlist.
What changed
Buyers ask AI now. The claims in its answer decide the shortlist.
More buyers skip the search results and ask ChatGPT, Perplexity, Gemini or Google's AI Mode instead. The answer they get is a set of claims about each vendor: "easiest to set up", "gets expensive past 50 seats", "best for small teams". Those claims are pieced together from a handful of sources, and Reddit threads and LinkedIn posts are among the ones these tools cite most.
Being mentioned is only the start. If the answer names you with a competitor's framing, or drops the claim that sets you apart, buyers hear the wrong reason to pick you.
Our job is to find the claims AI makes about your category, trace where they come from, and help you own the ones that decide the shortlist.
Search era
- Ten blue links
- The buyer picks which site to trust
- Goal: rank on page one
- Work on your own pages
Retrieval era
- One written answer
- The AI picks for them
- Goal: the answer states your claims
- Work on every page the AI reads
+527%
Growth in AI-referred sessions, year over year, first five months of 2025 (Previsible data reported by Search Engine Land)
40.1%
Share of LLM citations that point to Reddit (Semrush, 150,000+ citations)
11%
Share of domains cited by ChatGPT that Perplexity also cites (Profound, 680M citations), so claims differ by engine
Where budgets leak
Three common playbooks that miss AI answers.
Each one keeps a team busy. None of them, on its own, gets you named when a buyer asks AI. You can test each one yourself in ten minutes.
People see the ad, then forget it.
The pattern
Paid social, display, influencers. Judged on reach, CPM, and brand recall.
Why it breaks
An ad leaves nothing behind for an AI to read. When the buyer asks ChatGPT a month later, the answer is built from other people's content.
Test for it
Ask five AI tools: "Who are the leading vendors for [the job your buyers need done]?" If you never come up, your awareness spend isn't reaching the answer.
Paid search stops working the day you stop paying.
You rank, but you don't get quoted.
The pattern
Blog libraries, programmatic pages, keyword targeting, schema, internal links.
Why it breaks
Ranking a page and being quoted in an answer are different things. AI tools trust claims they can see repeated in several places. Your own site saying you're great counts for the least.
Test for it
Pick ten questions your best customers ask and check who AI quotes. If your own site is the only place you show up, that's a shrinking channel.
40.1% of AI citations point to Reddit. Company sites barely register.
Real trust, in places AI can't read.
The pattern
LinkedIn, X, Instagram, Discord, Slack groups, paid newsletters, founder content.
Why it breaks
Much of social is hard for AI to read or locked away completely. Most LinkedIn posts never get quoted, and Discord and Slack are invisible.
Test for it
Take your ten best social posts and check whether any show up in AI answers on related questions. If none do, that goodwill isn't helping you there.
About 90% of any community only reads. The people commenting are a small slice.
How AI answers work
An AI answer is a handful of claims. Whoever shapes them wins the shortlist.
AI assistants compress what's been said about your category into a few statements a buyer can act on. Each statement was formed somewhere: a thread, a review, a comparison page, your own site. We map those claims, trace where they came from and help you own the ones that decide who gets picked.
A buyer asks
"What's the best expense software for a 40-person startup?"
The AI answers
For a 40-person startup, Brexa is the budget pick: it's free for teams under 501. Ledgerly is the fastest to set up, at about 10 minutes2, and it connects to QuickBooks and Xero. Tallyfold is described as secure3, and its policy controls scale from 10 to 1,000 employees4. When choosing, weigh cost at your size, setup time, accounting integrations and how policies hold up as you hire.
Example with fictional companies.
- 1Brexa is the budget pickCompetitor's framing
Sets cost as a buying criterion
- 2Ledgerly is fastest to set upCompetitor's framing
Tallyfold claims setup in under an hour. The sharper number won.
- 3Tallyfold is secureTallyfold's claim
Survived, but the SOC 2 and SSO detail was dropped
- 4Tallyfold scales from 10 to 1,000 employeesTallyfold's claim
Survived word for word because it is a concrete range
How it works
Four steps. The first happens once.
We audit once. Then we plan (Architect), do the work (Deploy), and keep it current (Maintain), every month. This isn't a campaign with an end date. It's closer to keeping your reputation in good shape.
We run your buyers' questions through five AI engines, record every claim the answers make about you and your competitors, and trace the sources most likely behind each one. Delivered as the Category Claim Map, this is the only phase with an end date.
What happens:
- •Every claim AI makes about you and 3 to 5 competitors
- •The buying criteria AI names, and who owns each one
- •A map of which threads AI actually cites for your questions: Reddit and forums, LinkedIn, reviews, lists, your site, entity data
- •A check on which of your existing content and mentions are going stale
- •Three target claims: one to defend, one to shift, one to establish
You'll know which claims decide your category and which ones you can move.
We pick the threads and communities worth showing up in, brief the people on your team who will post, and write your key claims so they still read correctly when an AI shortens them to a sentence. We also agree the research we'll publish.
What happens:
- •Which threads to work, by type, and the rules for each: who judges it and what gets removed there
- •The phrases and definitions you want buyers to associate with you
- •Key claims tested at long, short, and one-line length to make sure they survive
- •A plan for original research published under your name
- •Tone and rules for each community
Nothing gets posted without a plan for how people and AI will read it.
Ongoing, genuine participation in the places AI reads. Original research each quarter. Before anything goes live, we check how AI tools summarize it.
What happens:
- •Work in each chosen thread, done its way: disclosed posts on Reddit, named voices on LinkedIn, evidence for editors, clear answers on your site
- •Original research, signed and dated
- •Supporting pieces: explainers, comparisons, newsletters
- •A pre-publish check of how four AI tools summarize each key claim
- •Getting the same facts repeated on other trusted sites
Content that keeps getting quoted for months, not days.
Sources lose weight over time. A great thread from 2024 can lose out to an average but busier one in 2026. Keeping things current is part of the job, not an extra.
What happens:
- •A monthly report on how often you're named
- •A quarterly check on every source, paced by how fast each thread type fades (Reddit fastest, entity data slowest)
- •Reviving threads when your mentions start to slip
- •Dated updates so there's always recent proof of the same claim
- •Watching competitors and responding when they move
Your position builds each quarter instead of fading once a campaign ends.
What We Won't Promise
There's no shortcut here, so we won't pretend there is.
If you need this quarter's pipeline from one channel, we're not the right fit. If you want AI to describe your category the way you do, read on.
Fit
Who this is for.
We work best with a certain kind of company, and we turn down a lot of enquiries. Here's how to tell if you're one of them.
Strong fit
Poor fit
What we measure
Four numbers that show whether AI is telling your story
Mention counts tell you whether AI named you. These tell you what it said, whose framing it used and what you can change.
Claim survival
When AI summarizes us, what's left?
Each of your claims is scored kept exact, paraphrased, distorted or dropped in the AI's answer, then combined into a score out of 100.
Claim adoption
Is AI repeating our framing?
The share of answers to your buyer questions that state a target claim in substance, by engine, on a four-week rolling average.
Criteria ownership
Who decides how buyers judge the category?
Of the buying criteria AI answers name, the share that match your framing versus each competitor's.
Fragility
Which claims can we move?
How few sources hold a claim up, how old they are and how much it varies between runs. Fragile claims are the cheapest to change.
AI answers change from run to run, so every question runs several times per engine and adoption is reported as a four-week rolling average. No one can guarantee what an AI engine says.
What you end up with
After 6 to 12 months of steady work, you should have:
- →AI answers describe you the way you'd describe yourself to a buyer
- →The buying criteria AI names are ones you win on
- →Evidence on your site that AI repeats with your numbers intact
- →Your experts showing up in the threads AI cites for your category
- →An early warning when a new claim about you starts forming
Once AI adopts a claim, it keeps repeating it while the sources behind it stay current.
Every report shows which of your claims AI now repeats, by engine.
Before you book a call
Fit, cost, timing, and how we measure results.
These come up on almost every first call. Answering them here means we can spend the call on your business.
Companies with a clear category, known competitors, and the patience to give it 6 to 18 months. You already have customers and want to be the name AI gives when a buyer asks, rather than chasing a pipeline number this quarter. If you need quick leads or big reach numbers, we're not the right team.
See what AI keeps from your positioning
Paste your homepage copy and a buyer question. In two minutes you'll see which of your claims survive when AI compares you with competitors, and which get lost.