Skip to content

AI Search Growth · B2B companies

We make your company discoverable in Google, ChatGPT and AI search.

We help companies become the answer when customers ask AI what to buy.

We are not an SEO agency with a new adjective. We measure whether answer engines name you, fix the reasons they do not, and report the result all the way through to pipeline.

Generated answersample

“Best payment platform for SaaS companies in Australia?”

  1. 1Northgate Pay
  2. 2Payflow
  3. 3Lumen Payyou
  4. 4Veritas Pay
Sourceslumenpay.example/guides/payment-platform-for-saasfintechreview.example/payment-orchestration-2026reddit.com/r/SaaS

Why this matters

Buyers stopped making their own shortlists

A B2B buyer used to open ten tabs. Now they ask an assistant which vendors to consider, and start from the four names it gives them. If you are not one of the four, you are not in the evaluation — and you will never see the traffic you did not get.

The shortlist moved upstream

Ranking fifth for a keyword still put you on the page. Being the fifth-best-documented vendor in a category gets you left out of a four-item answer entirely. Generative answers do not have a page two.

You cannot fix what you do not measure

Most companies have no idea whether ChatGPT names them, how often, in what position, or which competitor it prefers. That is a measurable thing, and measuring it is where we start.

It is winnable right now

In most B2B categories almost nobody is doing this deliberately. The companies that get named are usually the ones that happen to have good comparison content and third-party coverage. That is reproducible on purpose.

What GEO actually is

Generative Engine Optimisation is not a new set of tricks

It is the work of making your company understandable, quotable and corroborated. Answer engines have to retrieve something, ground their answer in it, and decide you are credible. Each of those is a separate problem.

Retrieval
Something has to be findable and parseable. If your integrations page needs JavaScript to show content, some retrieval pipelines see an empty page. This is old-fashioned technical SEO, and it still gates everything.
Grounding
The model needs a passage it can turn into a sentence. Pages that answer a question directly, name alternatives explicitly, and state specific numbers give it something to work with. Pages that only describe your product do not.
Corroboration
Your own claim about yourself is weak evidence. Independent sources — publications, review platforms, communities, expert coverage — are what turn a candidate into a recommendation. And the set of sources that matters is measurable: we can see which domains the engines actually cite in your category.
Entity clarity
The model has to be sure which company you are. Inconsistent names, descriptions and profiles across the web make you harder to resolve, and easier to skip.
SEO

Get found

Machines can crawl your site, parse it, and understand what your company is and does.

Content

Get cited

There is something on your site worth quoting, on pages that answer the questions buyers actually ask.

Authority

Get chosen

Independent sources corroborate it, on the domains these engines already trust.

AI searchWebsiteLeadCustomerRevenue

What you get

A measurement, not an opinion

Every engagement starts by probing your category across ChatGPT, Gemini, Perplexity and Google's AI surfaces, three repeats per query, and putting your competitors on the same chart as you.

Sample data. Everything shown here is built on a fictional company and fabricated probes. It shows the format of the deliverable, not a client result. The metric calculations behind it are real; the inputs are invented.

AI Search Visibility

Lumen PaySample data
Mention rate
35.0%
14 of 40 queries · +17.5pp vs baseline
Citation rate
22.5%
Your pages used as a source · +17.5pp vs baseline
Avg. position
4.2
When named in a ranked answer
Share of visibility
10.0%
Of all vendor mentions across 6 tracked brands

Six-month trend

Sample data
0%13%25%38%50%040506070809
Mention rateCitation rate
33/ 100
Composite score

Weights are published: 40% mention rate, 35% citation rate, 15% position, 10% share. Never reported without its components.

Share of AI visibility

40 queries · 4 enginesSample data
Lumen Pay10.0%pos 4.2
Northgate Pay27.1%pos 1.2
Payflow22.9%pos 2.3
Veritas Pay20.0%pos 2.6
Kestrel Financial10.7%pos 3.3
Obsidian Payments9.3%pos 3.5

Share of every vendor mention across the probe set. Competitor names are fictional.

By answer engine

Sample data
ChatGPTMentions41.7%5/12Citations25.0%Avg. position5.0
GeminiMentions30.0%3/10Citations20.0%Avg. position4.7
PerplexityMentions50.0%5/10Citations30.0%Avg. position3.0
Google AIMentions12.5%1/8Citations12.5%Avg. position—

Reported per engine rather than averaged. An average across engines hides the disagreement that matters — here, one engine names Lumen Pay half the time and another almost never.

The recurring engagement

Then the same loop, every month

AI visibility is not a project you finish. Queries change, competitors publish, models update. The retainer exists because the work is continuous — not because we invented monthly tasks to justify one.

01

Monitor

  • Google organic visibility
  • AI visibility across four engines
  • Competitor visibility
  • Backlinks & brand mentions
  • Community discussions
  • Reviews & media
  • New relevant queries
02

Identify

  • New search questions
  • Content gaps
  • Competitor advantages
  • Missing structured data
  • Citation opportunities
  • Authority & PR opportunities
  • Technical issues
03

Execute

  • Content creation & updates
  • Technical SEO improvements
  • Structured data
  • Internal linking
  • Digital PR
  • Authority building
  • Review strategy
04

Measure

  • AI mention rate
  • AI citation rate
  • Average AI position
  • Share of AI visibility
  • Google rankings & traffic
  • Organic conversions
  • Referring domains

Why this is defensible

Agents research. Humans decide.

The research, analysis, drafting and reporting are systematised. The strategy, the expertise, the relationships and the approvals are not. We do not publish anything off-site or contact anyone without a named human approving it.

Agents do the research

  • Crawling and technical analysis
  • Competitor analysis
  • AI query generation and probing
  • Citation and mention tracking
  • Question and content research
  • Community and PR opportunity detection
  • Briefs, drafts and outreach drafts
  • Reporting

Humans decide

  • Strategy and prioritisation
  • Subject-matter expertise
  • Approval of anything published or sent
  • Relationships with press and communities
  • Publishing decisions
  • Client communication
  • Judgement calls
  • Interpretation of results

From findings to revenue

Prioritised actions, then the numbers that pay for them

Findings are worthless without an order of operations, and visibility is worthless if it never reaches pipeline. Both are part of the deliverable.

Sample data. Everything shown here is built on a fictional company and fabricated probes. It shows the format of the deliverable, not a client result. The metric calculations behind it are real; the inputs are invented.

Recommendation engine

impact x confidence / effortSample data
  1. 01

    Fix Organization schema and align entity data across external profiles

    Your Organization markup omits sameAs, founders and a canonical description, and three external profiles describe the company differently. Inconsistent entity data makes it harder for a model to resolve you as one company.

    seoimpact higheffort lowconfidence highEngineering
    7days 0-30
  2. 02

    Make pricing information crawlable and specific

    Pricing sits behind a form. Every pricing query in the probe set was answered using third-party estimates, which were wrong. Publishing a real pricing explainer replaces a guess with your own words.

    contentimpact higheffort lowconfidence highProduct marketing
    7days 0-30
  3. 03

    Instrument AI referral traffic and connect it to pipeline

    Visibility that cannot be traced to sessions and leads will not survive a budget review. Instrumentation must precede the growth claims.

    growthimpact higheffort lowconfidence highGrowth + engineering
    7days 0-30
  4. 04

    Publish a genuine comparison hub for the three queries you lose most

    Answer engines pull shortlists from pages that compare options explicitly. You have no page that names an alternative, so there is nothing to quote when a buyer asks for a shortlist.

    contentimpact criticaleffort mediumconfidence highContent lead + product marketing
    5.56days 0-30
  5. 05

    Add internal links from the blog into the six pages that can win answers

    Your highest-intent pages sit four clicks deep with almost no internal links. Depth suppresses both crawling and perceived importance.

    seoimpact mediumeffort lowconfidence highContent lead
    4days 0-30

The priority score is published, not proprietary: impact x confidence, divided by effort. A client who disagrees with the ordering can argue with the inputs, which is the point.

Visibility to revenue

Month over monthSample data
  1. AI answers naming you14of 40 queries
  2. AI answers citing your site9of 40 queries
  3. AI referral sessions486sessions / month
  4. Organic sessions8,420sessions / month
  5. Qualified leads37leads / month
  6. Pipeline created148,000USD / month

Nobody buys “you appear 12% more often in AI answers”. The chain has to end in pipeline, and the parts of it that are estimates rather than measurements are labelled as estimates in the real report.

How we work

What we will not do

This matters more in this discipline than in most, because the shortcuts are obvious and they work for a while.

  1. 01

    No fake accounts, fake reviews or astroturfing

    No Reddit account farms, no reviews we wrote, no sockpuppets. Our systems find opportunities; real people with real names participate. Platform manipulation has a short half-life and our clients are regulated companies who cannot afford it.

  2. 02

    No hundreds of thin AI-written pages

    Programmatic content at scale is the fastest way to look like every other vendor and get treated accordingly. We build a smaller number of pages that answer real questions, sourced from interviews with your team.

  3. 03

    No invented numbers, ours or yours

    No fake testimonials, no borrowed client logos, no case-study statistics we cannot reproduce. Every number in your report traces back to a stored measurement with a model version and a timestamp.

  4. 04

    No black-box scoring

    Our composite score, priority formula and module weights are published. If you disagree with our ordering, you can argue with the inputs.

  5. 05

    No pretending the measurement is more precise than it is

    Generative answers are non-deterministic. We probe repeatedly, report ranges where ranges are honest, and tell you when a month's movement is noise.

Next step

Find out where you stand in AI search

The audit starts with a measurement: 40 queries, four answer engines, your competitors alongside you. You get the numbers, the reasons, and a 90-day plan.