We measure whether ChatGPT, Perplexity, Gemini and AI Overviews recommend your brand — using a prompt panel we publish, with error bars — and then we do the work that changes the answer.
One real prompt, one real answer, no charge. Reply in two working days.
Answer-engine optimization (AEO), also called generative engine optimization (GEO), is the practice of getting a brand named and cited inside AI-generated answers from systems like ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude. It differs from SEO because there are no ranked results to climb: an answer engine names a handful of options and omits everything else. The objective is inclusion in that shortlist, not position on a page.
That single structural difference changes what the work is. You cannot inch from position eleven to position seven. You are either in the answer or you were never in the deal.
A ranking puts you at position eleven. An AI answer does something worse — it names three vendors and stops. There is no page two.
A buyer types "best X for teams under 50" and gets three names with trade-offs already summarised. By the time anyone visits a website, the field has been cut — by a model, using sources you do not control.
Models cite what other trusted sources say about you far more than what you say about yourself. A first-position ranking and a strong domain do not guarantee a mention. Plenty of category leaders are losing to smaller, better-cited rivals.
It is not SEO, not PR, not content, and not paid. It borrows from all four, which is exactly why it falls between them and why the gap widens quietly for months before anyone measures it.
Published benchmark scores for AI presence. These are not close races — and in each pair, both products are good. The difference is citability, not quality.
Prompt class: "best workflow automation tool for a small ops team"
Scores from a published 2026 third-party benchmark study of AI presence in B2B SaaS, measured across major answer engines. Reproduced with attribution; Shortlisted did not produce these figures. Your own score is measured against a prompt panel built for your category, not this one. More on the benchmark data →
Most AI visibility scores have a hidden denominator. The universe of possible prompts is infinite, vendors sample an arbitrary slice of it, then report a percentage as though it were a fact. We built our measurement around fixing that, in public.
A fixed panel of 120–200 buyer-intent prompts, written with you and signed off before a single measurement is taken. It is printed inside every report. We never quote a percentage without showing the prompt set that produced it.
No brand-name queries. Asking a model about a company it was just asked about proves nothing. The panel contains only what a buyer asks before they know you exist.
Every prompt runs at least five times per engine per cycle. Model outputs vary run to run, so one pass is noise wearing a suit. You get the mean and the variance band.
When a provider ships an update, scores move for reasons no agency caused. In our reports that appears as a labelled model event, not as your failure.
Recommendation Rate — how often you are named when a prompt asks for a shortlist. Citation Share — how often your domain is the linked source. Narrative Position — how you are framed when named.
Tell us your brand and category. We run a real buyer prompt against live answer engines and send you exactly what came back — who was named, who was not, and the source each model leaned on.
2 working daysThe complete panel across four engines with variance bands, a competitor leaderboard, all three metrics, and a ranked list of the ten specific reasons you are losing — each one costed and sequenced.
10 working daysMonthly panel re-runs, third-party placement, entity and content fixes, and one strategy call a month. You watch the Recommendation Rate move, with the error bars left in.
3-month minimumEvery price we charge is on this site. If a vendor needs a discovery call before telling you a number, that is a negotiating position, not a scoping requirement.
The diagnosis. Buy it, act on it yourself, and never speak to us again if you like.
The default engagement. Measurement plus the work that changes the measurement.
For brands that intend to own the shortlist rather than appear on it.
We will not go below $3,000 a month, and you should be suspicious of anyone who will. Below that price you are buying a dashboard. Dashboards are useful and cheap — a licence runs $99 to $400 a month — but a dashboard has never earned a citation, placed an article or fixed an entity. If all you need is measurement, buy the tool and keep your money.
You should, if measurement is all you want. Those are good products and we use tools like them in delivery — the licence is inside our fee.
The distinction is simple: a dashboard tells you that you are losing. It does not write the comparison article you are missing from, negotiate the third-party placement, fix the entity data confusing the model, or run the community presence that shifts retrieval. That work is people, and people are what you are paying for.
No, and we will not claim it. Anyone who promises you traceable revenue attribution from AI answers today is selling you something they cannot deliver.
What we commit to is movement in Recommendation Rate — how often you are named when a buyer asks for a shortlist — measured against a panel you approved, with the variance shown. We agree that success metric in writing before we start.
It will happen. When one provider shipped a major update in late 2025, citation counts moved across every dashboard in the industry for reasons unrelated to any brand's actual relevance.
Every data point we record carries its model version. When a discontinuity appears, it is labelled in your report as a model event rather than presented as your failure or our success.
First measurable movement in Recommendation Rate typically appears in the second or third monthly cycle. Citation Share moves more slowly because it depends on third-party publishing timelines we influence but do not control.
Anyone quoting you a faster timeline is either optimising brand-name prompts, which proves nothing, or guessing.
Never. Community presence genuinely matters to how models retrieve, and that makes the temptation real — which is exactly why the answer has to be absolute.
We disclose affiliation every time. A detected manipulation campaign would end your relationship with the community permanently and take your brand down with it.
Give us a brand and a category. We run a real buyer prompt against live answer engines and send you the raw answer — including the part where a competitor gets named and you do not.