- What goes wrong
- Buyers ask answer engines for product recommendations before they ever reach a category page. Most stores have strong product pages and nothing an engine can use in a ranked answer, because a product page describes one item to somebody who has already chosen the brand.
- What we change
- We make the brand resolvable as an entity, publish the comparison and shortlist pages nobody has written about you, and target the specific review and roundup domains the engine already cites in your category.
- What gets measured
- Citation rate on discovery, shortlist and attribute prompts, scored against three competitors on identical queries.
30-day AI visibility sprint
Get named by the engines your buyers ask.
We measure how often AI answer engines cite your brand, fix the three things that decide it, and re-scan on identical prompts thirty days later. Ecommerce, SaaS, Web3 and apps.
Start with the free check — it is the same baseline we work from, or tell us what you need.
How the sprint runs
Four phases, thirty days.
- 01
Baseline
Twenty prompts across six answer engines, five runs each, for you and three competitors. Every raw response is stored. You get a citation rate and a competitor gap that can be re-run and checked rather than taken on trust.
- 02
Diagnosis
We map every citation back to its source and separate the three causes: entity signals the engine cannot resolve, comparison questions nothing on your site answers, and third-party domains where you are absent and your competitors are not.
- 03
Execution
Structured data and entity fixes shipped, the comparison and shortlist pages written, and outreach aimed at the specific domains the baseline showed being cited instead of you. Prioritised by what the data says will move first.
- 04
Re-scan
The identical prompt set re-run at day thirty against the stored baseline. Before and after on the same queries, on the same engines, with the raw responses from both runs kept as evidence.
Who we do it for
One method. Four very different failure modes.
The measurement is identical across industries. What differs is the questions buyers ask, the competitors they weigh you against, and the source domains an engine trusts in the category.
- EcommerceDTC brands and online stores
- SaaSB2B software and tools
- Web3Protocols, wallets and infrastructure
- AppsMobile and consumer products
- What goes wrong
- Software buying starts with 'best tool for X', 'X vs Y' and 'alternatives to X'. Those are the three query shapes an answer engine handles best, and the three most SaaS sites refuse to write about honestly. The engine fills the gap from review platforms and competitor comparison pages instead.
- What we change
- We build the comparison and alternatives surface properly, restructure documentation into extractable answers, and fix the review-platform and directory signals that engines lean on for B2B software.
- What gets measured
- Citation rate on comparison and alternatives clusters, plus which competitor is named when yours is not.
- What goes wrong
- In crypto the failure is rarely silence — it is being described wrongly. Engines conflate protocols with similar names, repeat superseded documentation, and pull from sources that a serious user would never trust. Being cited inaccurately costs more than not being cited.
- What we change
- We disambiguate the entity across every surface an engine reads, get documentation current and machine-readable, and build corroboration on the sources that hold up to scrutiny in the category.
- What gets measured
- Citation rate and citation accuracy: whether the engine names you, and whether what it says about you is correct.
- What goes wrong
- App discovery has moved from store search to 'what's the best app for X'. Answer engines build those responses from editorial roundups, forum threads and store listings — three surfaces most app teams treat as marketing afterthoughts rather than as the retrieval layer they now are.
- What we change
- We fix listing and entity metadata, earn placement in the roundups the engine actually reads, and create the use-case pages that match how people phrase app questions.
- What gets measured
- Citation rate on use-case and shortlist prompts across the engines your audience actually uses.
What you get
Five artefacts, all re-checkable.
- 01Baseline report: 20 prompts × 6 engines, you and three competitors
- 02Entity and structured data specification, ready to implement
- 03Comparison and shortlist content plan, page by page
- 04Ranked target list of the source domains cited instead of you
- 05Day-30 re-scan on identical prompts, with the raw responses from both runs
Questions
Before you email.
Three things move a citation rate: structured data that lets an engine resolve the brand as an entity, pages that answer comparison and shortlist questions directly, and mentions on the third-party domains the engine already cites in the category. A sprint ships all three, prioritised by what the baseline scan shows is costing you the most citations.
An SEO retainer optimises for position in a ranked list of links and reports on rankings and clicks. This work optimises for being named inside a generated answer, and reports a citation frequency measured across repeated runs on multiple engines. The underlying craft overlaps; the target and the scoring do not.
Ecommerce and DTC brands, B2B SaaS, Web3 protocols and infrastructure, and consumer apps. The measurement method is identical across all four; the prompts, the competitor set and the source domains that matter are specific to each.
Changes that live on your own pages typically surface in answers within two to six weeks as engine indexes refresh. Changes that depend on third-party sources take longer, because they only register once the source publishes and is itself re-crawled. A sprint is scoped to thirty days so the on-site work is measurable inside it, with the third-party effects following.
Every scan stores the raw response body, the prompt, the engine and the timestamp. The day-30 re-scan uses the identical prompt set against the stored baseline, so the before and after are comparable rather than anecdotal. If the number did not move, the stored responses show that too.
Talk to us
Tell us what you are losing.
One reply, from the person who would do the work. If you have already run the free check, mention the domain and we will start from that baseline instead of asking you to explain it.