The offer, in plain terms
We measure how often AI assistants name and cite your brand, work out why you lose the prompts you lose, and do the work to change it. Monthly reporting with defined metrics, a named owner, and a work queue your team can see.
We cannot guarantee that an assistant will recommend you. Nobody can, and anyone who says otherwise is describing a system they do not control. What we can commit to is a measured baseline, a specific list of gaps, and the execution to close them.
At a glance
What you get | Monthly measurement, a prioritised work queue, and the content and technical execution to clear it |
What we measure | Mention rate, citation rate and recommendation rate across your prompt set, each defined and reported with sample size and range |
Typical engagement | Quarterly, with a 90-day initial programme |
Who runs it | A named strategist, with content and technical specialists |
Who it suits | Teams with a real content or measurement constraint and someone internally to work with |
Who it does not suit | Teams looking for guaranteed placements, or who have not fixed their free foundations yet |
The short version of our method
Measure a frozen prompt set properly. Diagnose each loss into one of four causes. Fix the cause. Re-measure against the same set. Report what moved and what did not.
Related reading: AI Visibility Optimization: 2026 Guide
That is less exciting than most pitches in this category and it is the part that actually works.
Anchor targets assume your CMS generates slug IDs from heading text. Confirm on staging and adjust the hrefs if your platform slugifies differently.
The problem this solves
Buyers increasingly ask an assistant before they search, and assistants name a few options rather than listing ten. If you are not among those options you are not in the consideration set, and you will not see it happen in your analytics because there is no impression to count.
That is the problem. It is real and it is narrower than most marketing on this subject suggests.
What is actually changing
Claim you will see | What the evidence supports |
|---|---|
Organic traffic is collapsing | Click-through is measurably lower on queries with AI summaries. A Pew study of 68,879 searches found 8% organic click rate with an AI summary present against 15% without |
Nobody clicks any more | AI referral traffic remains a small share of sessions. Conductor's November 2025 analysis across 13,770 domains put it around 1.08% |
AI traffic is worthless | It converts well. Semrush reported AI-referred visitors converting at around 4.4 times the organic rate |
The honest summary: this is a small, high-quality channel growing from a small base, alongside a measurable decline in clicks on informational queries. Both facts matter and most vendors quote only one.
Why it is hard to handle internally
Not because it is mysterious. The mechanics are knowable and we publish them openly. It is hard because it needs three things most teams cannot assemble at once:
Requirement | Why it is scarce |
|---|---|
Disciplined measurement | Frozen prompt sets, multiple runs, variance reported. Tedious and skipped |
Content production at a reliable cadence | The constraint for most teams |
Technical diagnosis | Crawler access and rendering sit with engineering, whose backlog you do not own |
A team with all three does not need us. Most teams have one.
What we are not claiming
That traditional search is finished, that you must act immediately or lose your market, or that we can place you in an answer. The pressure in this category is largely manufactured, and a service sold on urgency tends to be bought on urgency and judged harshly later.
This is worth doing because the work compounds and the foundations are cheap. That is a better reason than fear.
Who this is for
A good fit
You | Why it works |
|---|---|
Have a named person we can work with weekly | The work needs an internal counterpart who can get decisions made |
Publish, or want to publish, four or more pieces monthly | Below that the production half has little to do |
Can review content for factual accuracy | We write; your people verify claims about your product |
Compete in a category where buyers research before buying | B2B software, professional services, considered purchases |
Want measurement you can defend internally | Our reporting is built to survive a finance review |
Not a good fit
We would rather say this now than three months in.
You | What to do instead |
|---|---|
Want guaranteed placement in AI answers | No provider can deliver that. If one promises it, ask how, in writing |
Have not completed the free foundations | Finish your Google Business Profile, directory consistency and Search Console setup first. It costs an afternoon and we would only tell you to do it |
Are a single-location local business | Your highest-return work is profile completeness and reviews. That is free and you do not need us |
Have blocked AI crawlers and will not revisit it | If retrieval crawlers cannot reach your pages, nothing downstream works. That decision comes first |
Want link building | We do not do it. See the next section |
Need results this quarter to save a number | Realistic timelines are in the results section. This is not a rescue |
Have nobody to review content | In regulated categories especially, unreviewed content is a liability rather than an asset |
The honest filter
If your foundations are unfinished, the most valuable thing we can tell you is to go and finish them. We will say so in the first conversation, and it costs you nothing.
About a third of the audits we run end with us recommending work the client can do themselves. That is a reasonable outcome and we would rather be the people who said it.
Scale
Organisation | Fit |
|---|---|
Solo or very small, under 4 pieces monthly | Not yet |
Small team with a production constraint | Good fit |
Mid-market with several product lines | Good fit |
Enterprise, multiple brands and markets | Good fit, with a different scope and longer timelines |
Agency serving clients | Talk to us about white-label arrangements |
What is included
Four workstreams. Most engagements weight them differently depending on where your constraint sits, and we tell you which after the audit rather than before.
1. Measurement
Deliverable | Detail |
|---|---|
Prompt set construction | Built from your sales calls, support tickets and Search Console data, split across category, comparative, problem-first and branded intents |
Versioned and frozen | So comparisons across months mean something |
Monthly sampling | Multiple runs per prompt across the assistants your buyers use |
Three metrics reported separately | Mention rate, citation rate, recommendation rate, each with sample size and range |
Competitor set tracking | Who is named instead, and which of their pages were cited |
Full answer archive | Exportable, yours, retained |
2. Diagnosis
Every lost prompt sorted into one of four causes, because each has a different fix and a different owner.
Cause | What it means |
|---|---|
No coverage | The question is unanswered on your site |
Covered but not extractable | You have a page; a competitor's passage was quoted |
Not retrieved | Crawler access, indexing or rendering |
Described inaccurately | The source needs correcting, on your site or a third party |
3. Execution
Workstream | What we do |
|---|---|
Content production | Research, planning, drafting and publishing against the gaps the diagnosis found |
Page restructuring | Rewriting openers as direct answers, splitting near-duplicates, adding fit and non-fit statements |
Technical fixes | Crawler access, schema, rendering. We produce the specification and work with your engineering team |
Entity consistency | Aligning your business and product facts across the sources assistants read |
4. Reporting and review
A monthly pack with the metrics, what shipped, what moved, what is blocked and with whom, and the decisions we need from you. A quarterly review that revisits the prompt set itself.
What you provide
From you | Why |
|---|---|
A named counterpart, roughly two hours weekly | Decisions and context |
Factual review of content before publication | You know your product; we do not, at the start |
Access to Search Console and server logs | Diagnosis depends on your own data |
A route to engineering for technical fixes | We specify; someone has to deploy |
That last row is where engagements most often stall. If technical changes cannot ship, say so early and we will weight the programme toward what can.
What we do not do
Several tactics are common in this category and we do not offer them. The reasons are practical rather than moral: each creates exposure that lands on your domain and your brand rather than on ours.
We do not buy links
Google's link spam policy names buying or selling links that pass ranking credit, and link building conducted through automated programs or services, as link schemes. Rebranding the tactic as AI authority building does not change what it is.
Instead: we earn mentions through genuinely useful material, original data and inclusion in roundups on merit. Slower, and it does not put your domain at risk.
We do not arrange community mentions
You will see services offering to ensure your brand appears in natural-seeming discussions on Reddit, X and similar platforms. That is astroturfing. It breaches platform rules, and FTC guidance on endorsements requires disclosure of a material connection between an endorser and a brand.
Communities also detect it reliably, and the reputational cost when they do exceeds whatever visibility it bought.
Instead: your people participate openly with their affiliation stated, where that is genuinely useful.
We do not coach reviews
Asking customers to include particular words or themes in a review produces something that is no longer the customer's own account, which creates problems under FTC guidance on endorsements and typically breaches the review platform's own rules.
Instead: we help you build a consistent request process, asking every customer, without incentive or scripting. Volume and recency are what influence AI recommendations anyway, and both come from asking.
We do not promise placements
We cannot make an assistant recommend you. Nobody can. Anyone offering guaranteed inclusion is either describing something they do not control or planning one of the tactics above.
Instead: we commit to the measurement, the diagnosis and the execution, and we report honestly on what moved.
We do not generate volume without review
Google's position is that content produced at scale with little added value falls under scaled content abuse regardless of how it was made. More pages on the same topic also compete with each other for the same passage.
Instead: we publish against identified gaps, one question per page, with factual review before anything goes live.
Why this list is on a sales page
Because you will be comparing us with providers who do offer these things, and the comparison is easier if you know where the difference is.
If guaranteed placements or purchased mentions are what you want, we are not the right provider, and we would rather you find that out here than after signing.
How an engagement runs
Before you commit: the audit
A paid diagnostic, credited against the first quarter if you proceed. About two weeks.
We do | You get |
|---|---|
Build a starter prompt set and measure it | Your actual baseline, with sample size and range |
Check crawler access in your server logs | Whether assistants can reach you at all |
Sample your templates for extractability | Which page types have no quotable passage |
Audit entity consistency across sources | Where your facts disagree with each other |
Review your free foundations | What you should fix yourself before paying anyone |
You keep the audit either way. If it concludes you do not need us, it says so, and that happens often enough that we price the audit to stand alone.
Days 1 to 30
Week | What happens |
|---|---|
1 | Kickoff. Confirm the prompt set, freeze v1.0, agree metric definitions and reporting format |
1 | Technical findings handed to your engineering team as a specification |
2 | Full baseline measurement across all assistants |
3 | First diagnosis: every lost prompt sorted into its cause |
4 | Content plan agreed against the gaps, with your priorities applied |
Days 31 to 90
Period | What happens |
|---|---|
Weekly | Triage of new losses; content in production; your factual review |
Monthly | Re-measurement against the frozen set; reporting pack; decisions required |
Day 60 | Mid-point review. If something is not working we change it here rather than at renewal |
Day 90 | Full review: what shipped, what moved, what is blocked, what next quarter looks like |
Ongoing
Quarterly cycles. Each one re-examines the prompt set against fresh sales and support language, since the set ages and a stale instrument measures a market that has moved.
Who you work with
A named strategist who runs the engagement and attends your reviews, supported by content and technical specialists. You will know who they are and they will not change without notice.
How it ends
Quarterly terms with 30 days' notice. Your prompt set, your answer archive and your measurement history are exported to you in a machine-readable format on request, during or after the engagement.
That is in the contract rather than at our discretion. Your measurement history is the thing you cannot rebuild, and it should not be leverage.
What we report
Three metrics, each defined precisely enough that you could compute them yourself from the raw answers we give you.
The definitions
Metric | Definition |
|---|---|
Mention rate | Runs in which your brand name appears anywhere in the answer, divided by total runs |
Citation rate | Runs in which a URL on your domain appears in the answer's sources, divided by total runs |
Recommendation rate | Runs where your brand is presented as a suggested option rather than referenced in passing, divided by total runs, using a written classification rule we share with you |
They are reported separately because they have different causes. Mentioned but not cited means engines know you from third parties and find nothing quotable on your site. Cited but not recommended means you are useful as a source and not seen as an option. Those need different fixes.
How a figure is presented
Citation rate: 23%
Prompt set v1.2 (120 prompts), 5 runs each, 600 runs total
Measured 1 to 7 October 2026
Assistants: ChatGPT, Gemini, Perplexity
Range across runs: 19% to 27%The range is there so that a move from 23% to 25% is not presented as progress. If a change falls inside your observed variance, we say so.
The rest of the pack
Section | Contents |
|---|---|
Prompt-level movement | Which specific prompts changed state, which matters more than the headline rate |
Competitive view | Who was named instead, and which of their pages were cited |
Work shipped | What we published and changed, with dates |
Work blocked | What is waiting, with whom, and since when |
Decisions needed | Named, with the person who has to make each |
The blocked and decisions sections are what make the monthly meeting useful. Metrics alone give a senior stakeholder nothing to act on.
On attribution
You will be asked what revenue this produced. The accurate answer is that AI referral attribution is incomplete: many assistants pass no referrer, many answers produce no click, and influence often appears as better-informed buyers rather than tracked sessions.
We say that plainly in every report, then give you what is measurable: citation rate on commercially relevant prompts, self-reported source data if you collect it at intake, and branded search volume as a secondary signal.
We will not model a revenue figure from citation rate. The assumptions required to do it are not defensible, and a number that collapses under scrutiny is worse than no number.
What results look like
The honest version, including the parts that are slow.
Timelines by work type
Work | When it shows up |
|---|---|
Crawler access fix | Days to weeks. The fastest change available, and sometimes dramatic |
Entity and directory consistency | Weeks. Affects how you are described almost immediately once sources agree |
Opener rewrites on existing pages | Weeks to a month. The page is already indexed |
New content against gaps | One to three months per piece before it is retrieved reliably |
Third-party corroboration | Quarters. Genuine coverage cannot be rushed |
Overall programme | A fair assessment at six months, not three |
The first quarter
Expect the first quarter to produce foundations rather than headline movement. Measurement in place, technical issues identified and specified, entity data aligned, the content plan running.
If your crawler access was blocked, the first quarter can produce a large change, because you were previously invisible for a fixable reason. That is the exception rather than the plan.
What we will not promise
Promise you may hear | Why we do not make it |
|---|---|
A specific citation rate by a date | Depends on your category's competitiveness, which we do not control |
Placement in specific answers | No mechanism exists to deliver it |
A traffic multiple | The arithmetic behind those figures is rarely shown |
Results by next quarter | Occasionally true, not plannable |
What we do commit to
Commitment | Detail |
|---|---|
A measured baseline in the first month | With sample size and range |
A diagnosis for every lost prompt | Sorted into its cause with an owner |
An agreed volume of execution each month | Specified in your contract |
Honest reporting | Including months where nothing moved, and saying why |
Your data exported on request | Machine-readable, during or after the engagement |
On case studies
We publish results only where we can state the sample, the period, the baseline, the method and the confounding factors. That standard means we publish fewer case studies than competitors do.
If you want to speak to a current client about how the engagement actually runs, ask, and we will arrange it. A conversation with someone doing the work is worth more than a number on a page with no methodology behind it.
Pricing
Placeholder figures. Replace with your actual tiers before publishing, or remove the numbers and keep the structure with a request-a-quote route.
What drives the price
Factor | Effect |
|---|---|
Prompt set size | More prompts, more sampling, more triage |
Number of assistants tracked | Each adds measurement cost |
Content volume per month | Usually the largest component |
Brands, markets or locations | Multiplies most of the above |
Technical scope | Whether we specify only, or also implement |
Content volume is the main driver for most engagements. Measurement is a relatively fixed cost; execution scales with what you want produced.
Indicative structure
Tier | Suits | Includes | Monthly |
|---|---|---|---|
Audit | Anyone deciding | One-off diagnostic, credited against the first quarter | One-off fee |
Core | Single brand, one market | Measurement, diagnosis, a set content volume, monthly reporting | From $X |
Growth | Several product lines | Core plus higher volume and technical implementation | From $Y |
Enterprise | Multiple brands or markets | Scoped individually, with workspace separation and governance | Quoted |
What is not charged extra
Included | Why |
|---|---|
Data export | Your measurement history is yours |
Reporting | Not a premium feature |
Your named strategist's time in reviews | Part of the engagement |
Prompt set revisions at quarterly boundaries | Keeping the instrument current is the job |
Terms
Quarterly, 30 days' notice, no automatic annual rollover. We would rather you stay because the reporting shows it is working.
How we compare on cost
An enterprise AI visibility tool alone typically runs four figures monthly before anyone acts on its output, and execution labour runs well beyond that at a loaded hourly rate.
The honest framing is that a managed service is not cheaper than software; it is cheaper than software plus the people to act on it, provided you would otherwise need to hire. If you already have that capacity, buy the tool and keep the difference. We will tell you that in the first conversation if it applies.
How to start
Before you talk to us
Twenty minutes that will make the conversation better, and may make it unnecessary.
Ask ten questions your buyers ask, twice each, in ChatGPT, Gemini and Perplexity. Record whether you were named, how you were described, and who appeared instead.
Check your Google Business Profile is complete, if you serve a location.
Confirm Search Console is connected and look at the generative AI performance report.
If that exercise shows your foundations are not in place, fix them first. It is free, it takes an afternoon, and it is the first thing we would tell you.
The first conversation
Thirty minutes. We will ask what your constraint is, what you have already tried, who internally would work with us, and whether technical changes can ship at your organisation.
We will tell you honestly whether we think we can help. Sometimes the answer is that you need a tool rather than a service, or a hire rather than either.
Then the audit
A paid diagnostic, about two weeks, credited against your first quarter if you proceed. You keep it either way, including the baseline, the technical findings and the entity audit.
If it concludes you do not need us, it will say so and explain what to do instead.
What we will ask you for
Item | Why |
|---|---|
Search Console access | Your own query data |
Server log access or a sample | The only definitive crawler check |
Ten to twenty real buyer questions | The starting prompt set |
Your main competitors | The comparison set |
Who internally would own this | The engagement needs a counterpart |
If you are not ready
That is a reasonable place to be. Our guides on free AI SEO tools and AI visibility measurement cover the work you can do yourself, in enough detail to actually do it.
Come back when you have a constraint you can name. The conversation is much shorter and the engagement works better.
FAQs
What is AI Discoverability as a Service?
A managed service that measures how often AI assistants name and cite your brand, diagnoses why you lose the prompts you lose, and executes the content and technical work to change it. Monthly reporting with defined metrics, a named strategist and a visible work queue.
Can you guarantee my brand appears in AI answers?
No, and no provider can. Assistants decide what to say using systems nobody outside the model providers controls. We commit to measurement, diagnosis and execution, and we report honestly on what moved. Treat any guaranteed placement offer as a reason to ask how, in writing.
Why not just buy a tool?
If you have people to act on what a tool surfaces, buy the tool. Acting on the findings is the expensive part, typically 25 to 70 hours monthly depending on volume. A service makes sense when you would otherwise need to hire for that capacity, and we will say so if it does not apply.
How is this different from SEO?
It shares the technical foundation, since being crawlable and indexed gates both. It differs in what is measured, which is citation in answers rather than position in a list, and in emphasis, since retrieval selects passages and favours direct answers and third-party corroboration.
Do you build backlinks?
No. Google's link spam policy names buying links and automated link building as link schemes, and rebranding that for AI search does not change it. We earn mentions through useful material, original data and inclusion on merit, which is slower and does not put your domain at risk.
How long before we see results?
Crawler fixes can show in days. Entity consistency in weeks. New content takes one to three months per piece to be retrieved reliably, and third-party corroboration takes quarters. Expect a fair assessment at six months rather than three.
What happens to our data if we leave?
Your prompt set, answer archive and measurement history are exported to you in a machine-readable format on request, during or after the engagement. That is contractual rather than discretionary, because your history is the one thing you cannot rebuild.
Can you fix incorrect information about our brand?
We can address the sources. There is no mechanism to submit a correction to a model, and nothing edits training data. What works is correcting your own outdated pages, fixing third-party listings, and publishing a clear current statement of the facts. It works through retrieval, over time, and not with certainty.
Start with the audit
Two weeks, credited against your first quarter if you proceed, and yours to keep either way.
You get your actual baseline with sample size and range, a definitive answer on whether AI crawlers can reach your pages, a template-level view of which page types have nothing quotable, and an entity audit showing where your facts disagree across sources.
If it concludes you should do the work yourself, it will say so and tell you what to do.
Book the audit or talk to us first if you would rather work out whether it is worth it.
Still deciding
Run the twenty-minute check yourself before you talk to anyone. Ten buyer questions, twice each, across three assistants, written down. That tells you whether you have a problem and roughly what kind, and it costs nothing.
Our guides on free AI SEO tools and measuring AI visibility cover the whole method in enough detail to run it without us.
We would rather you arrive with a constraint you can name than a worry you cannot.