AI Search Visibility
Get named.
Get cited.
Get shortlisted.
When a buyer asks an AI who they should be talking to, they get two or three companies and a reason to stop looking. This is the work of being one of them.
The stakes
The shortlist is built before you know the buyer exists
A buyer asks an AI who they should be talking to. It names two or three companies. They contact those companies.
You never see any of it happen. No impression, no click, no form fill, no lost-deal note to review later. The opportunity does not get lost. It never arrives.
This is the only part of the funnel you currently have no visibility into, and it now sits in front of everything else you spend money on.
What it costs you
What Happens When AI Can't Find You
of B2B buyers chose a different vendor because of what an AI told them. Not a worse pitch. A conversation you were never part of. G2, n=1,076, March 2026.
71%
now use AI chatbots somewhere in vendor research, up from 60% seven months earlier. This is not an emerging channel. It is where the shortlist is decided. G2, same study.
The Work
The work, in the order we do it
Some of this happens on your site. A surprising amount of it does not, which is why a monitoring subscription cannot solve it for you.
The technical work
The 42-Point Technical Audit
Every site we ship is tested against these before launch and again 30 days after. The thresholds are the real ones, not rounded up for a brochure.
TECHNICAL AUDIT | 42 CHECKS · RUN AT ONBOARDING, RE-RUN QUARTERLY
Agent access 11 Checks
- robots.txt directives for every named AI crawler
- WAF and CDN rules silently blocking agents
- Server response to non-browser user agents
- Server-side vs client-side rendering of primary content
- Response codes by agent: 200, 403, 404, 429
- Rate limiting that drops agent requests without logging
- llms.txt and llms-full.txt present and accurate
- Sitemap coverage and lastmod honesty
- Time to first byte for non-browser requests
- Redirect chains that agents abandon
- Cloudflare bot-fight mode misconfiguration
Entity and identity 10 Checks
- Organization schema with a complete sameAs graph
- Entity disambiguation where the name is a common word
- Wikidata presence and factual accuracy
- Knowledge panel and Business Profile consistency
- Person schema for named authors and experts
- Identical description across every directory
- Parent, subsidiary and product relationships declared
- Founding date, headcount, location, category stated
- Category self-description matched to buyer language
- Conflicting descriptions on legacy profiles
Retrieval structure 12 Checks
- Passage-level self-containment, models retrieve chunks
- Answer placement in the first two sentences
- Heading hierarchy mapped to real buyer questions
- Tables and lists a model can lift whole
- Explicit attribution and visible publication dates
- Semantic HTML rather than nested div soup
- Internal links that express topical relationships
- Canonical handling and near-duplicate consolidation
- Alt text written as description, not keywords
- Comparison and alternatives coverage present
- Specific claims rather than hedged paragraphs
- Content that survives being read without the page
Schema and off-site 9 Checks
- FAQPage, Article, Service, Product, Person, Organization
- Validation against Rich Results, not just presence
- Schema accuracy versus decorative markup
- Review platform profile completeness
- Structured data on third-party profiles you control
- Citation source health across referenced domains
- Author identity consistent across every publication
- Third-party descriptions matched to your own
- Broken or stale citations pointing at you
Crawlers we check for by name
4 in 5 websites fail the first group. It is the cheapest thing to fix and the one nobody checks.
Deliverables
What you actually get
The Baseline Weeks 1 to 14
Prompt set
30 to 50 real buying questions, built from your sales calls and lost-deal notes. You sign it off before anything runs.
Baseline run
Every prompt across 6 models. ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude. Verbatim answers captured, timestamped and stored.
Competitor map
Who gets named instead of you, how often, in what order, and the words the models use to describe them.
Source audit
Every URL the models are drawing from, ranked by how often it influences an answer. Usually the first genuine surprise.
Fix list
Ranked by effort against impact, each item costed and assigned. Yours to keep whether or not you continue.
Read-out
60 minutes, walked through by the person who did the work, not an account manager reading a deck.
The Work Ongoing
Re-run
The same prompt set, the same 6 models, measured against the fixed baseline. Movement is real or it is not reported.
On-site work
[X] technical and structural fixes implemented. Schema, entities, crawler access, page structure. We ship them, you do not receive a to-do list.
Content
[X] pages written and published. Comparison pages, alternatives pages and answer pages, in that priority order.
Off-site work
[X] citation targets pursued each month. Review profiles built or corrected, community presence, and placement in the sources the models quote.
Dashboard
Live, always current, your own logins. Not a PDF that arrives three days late.
Report
Every two weeks. What moved, what did not, what we shipped, what is queued, and what it cost. Plain language.
Review call
Every week. 45 minutes with the lead on your account, not an account manager.
Always on For the length of the engagement
Named lead
The same person from the first call to the last. No handover to a junior after the contract signs.
Direct channel
Shared Slack or Teams. Replies inside one business day.
Raw data
Every AI answer we have ever captured for you, stored and exportable, including the baseline recorded before you hired us.
Ownership
Content, data, accounts, profiles and access are yours. Nothing is held on our platform and nothing stops working if you leave.
How we report
Numbers you can reproduce
Every number we give you comes after the queries have run, never before. You can re-run any of them yourself and get the same picture.
Models are non-deterministic, so the same question returns different answers on different days. We run every prompt multiple times and report frequency, which is the only measure that holds up when you check it a week later.
Technical fixes move fastest and show up within weeks. Citation work follows once third-party sources are crawled and absorbed, which is why the reporting separates the two rather than blending them into one flattering line.
Timeline
When things move
Anyone promising results in days is describing one lucky prompt, not a position.
Common Questions
Frequently Asked Questions
How is this different from SEO?
SEO competes for a position in a list of links. This competes to be inside a single synthesised answer. The foundations overlap, which is why we run SEO as well, but the work of earning a citation is not the work of earning a ranking.
Which models do you cover?
ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Claude. Google’s AI surfaces matter most for volume, and ChatGPT sends the large majority of trackable standalone AI referral traffic.
Can you control what an AI says about my company?
Not control, no. You can change the inputs it draws from, which in practice changes the output. Anyone claiming control is selling something that does not exist.
What does it cost?
We quote after the free check, because until we have looked at your site we would be guessing. Engagements run on a 90-day starting term and we price the work, not a package.
Start Here
See what the answer is today
Enter your domain and watch it run. Eight real buying questions, three models, verbatim answers and the sources behind them. Under a minute.
Most free audits hand you a score. This hands you the actual answers. And if you want a person to walk you through what it found, we do that free as well. That part is not automated.
Under a minute · No sales call unless you ask for one
