Use case 03

The answer names your competitor. You are not in the shortlist it just wrote.

Someone asks an assistant which vendors to consider in your category. Four names come back, none of them yours. Worse, the buyer treats that list as a starting shortlist rather than a suggestion, so you are not losing the evaluation, you are not in it.

Talk to a strategist Free growth audit

The short answerBeing absent from AI answers is a retrieval and corroboration problem, not a volume problem. Assistants fan a question out into several searches, retrieve passages, and compose an answer from sources they can parse, attribute and trust. Pages that make unhedged, extractable claims about specific buying questions get cited. Generic thought leadership does not. We fix it by auditing the prompts your buyers actually use, building the comparison, pricing-model and limitation pages those prompts need, cleaning the entity and schema basics so pages can be retrieved, and earning third-party corroboration so the claims hold up.

The situation

How it gets described on the first call.

I asked ChatGPT who the vendors are in our space. We were not on the list.

CMO, enterprise software

Prospects arrive with a shortlist of three and we are explaining who we are.

VP Sales, cybersecurity

Our competitor is smaller than us and gets cited everywhere.

Head of Marketing, legal tech

We rank fine on Google. In the AI answer we do not exist.

Founder, supply chain software

Why the answer skips you even when you rank.

Google's own documentation describes the mechanism plainly: AI features in Search take a question, fan it out into multiple related searches, and compose an answer from what they retrieve. Google Search Central, AI features and your website, checked 4 October 2026. Google also states there are no extra requirements to appear, which is the important part. There is no AI schema to add and no vendor who can submit you. What gets you into an answer is being the clearest available source for the question the assistant fanned out to.

That is where most B2B sites fail. A model composing an answer needs a passage it can lift and attribute: a direct claim, in plain language, in a page section that stands on its own. Most vendor pages hedge every sentence, bury the claim three paragraphs down, or write for a reader who already knows the company. A page that says what it is, who it is for, what it costs to run, what it does not do, and how it compares to the named alternative gives a model something to quote. A page that says you are a leading innovative platform gives it nothing.

The second failure is corroboration. Assistants favor claims that appear in more than one place, especially places they already treat as reliable. If your only assertion of a capability is on your own site, it carries less weight than a competitor's claim that also appears in a review platform, a partner directory, a conference agenda and a practitioner's post. This is why AI visibility work is never only on-site work.

The third is simply absence. For most categories the buying prompts are specific: alternatives to a named incumbent, best tools for a named use case in a named industry, how a pricing model works, whether something integrates with a named system. If you have no page addressing a prompt, nothing can retrieve you for it. Across the audits we run, this is the single most common reason a company is missing, and it is also the fastest to fix.

Why the usual fix fails

What gets tried first. And why it does not hold.

AI visibility attracts more bad advice than any other channel right now. These four are the most expensive.

Publishing more blog posts

Volume aimed at informational questions adds pages nobody retrieves for a buying prompt. The engine was never short of content about your category. It was short of a quotable, specific answer from you.

Buying an AI visibility dashboard and stopping there

A tool reports your mention share across hundreds of prompts. Useful as instrumentation, useless as strategy. Nothing in the tool writes the comparison page that would change the answer.

Stuffing pages with questions and schema

FAQ blocks and markup help a machine understand a page that already says something. They cannot make a page that says nothing worth quoting into a citable source. Schema is hygiene, not leverage.

Chasing every engine at once

Separate programs for each assistant, each with its own report. Most of the work that earns a citation is platform agnostic: direct answers, self-contained sections, clean entity data and third-party corroboration. Splitting it multiplies cost without multiplying result.

The 80:20 cut here

The fifth of the work that moves this number. Twenty to forty prompts, answered properly.

The cut here is brutal and it works: a small prompt set that maps to real buying decisions, and pages built to be quoted for those prompts.

The cut for this situation

Where the first quarter goes. The right column is not wrong work, it is work that only pays once the left column exists.

Vital fifth

The twenty to forty prompts a buyer asks before a shortlist

Built from sales calls, search data and the questions your team answers weekly, not from a tool's suggestion list. These become the measurement panel and the content brief at the same time.

Vital fifth

Comparison, alternatives and limitation pages, written to be quoted

Self-contained sections, a direct claim in the first sentence, named alternatives, and an honest statement of where you are the wrong choice. The limitation page is the one competitors will not write, and it is the one that earns trust from both buyers and models.

Vital fifth

Entity and retrieval basics on the pages that matter

Consistent company facts everywhere you appear, Organization and Article schema, crawler access for the assistant bots, and page structure a model can segment. Hygiene, done once, on the pages that carry the claims.

Vital fifth

Third-party corroboration for the claims you care about

The same facts stated in review platforms, partner directories, conference agendas, podcasts and practitioner posts. One claim in five credible places outweighs five claims on your own domain.

What we park, and tell you we are parking

Tracking hundreds of prompts. Instrumentation inflation. Forty well-chosen prompts tell you more
A separate program per assistant. The work is mostly shared. Run one program, report per engine
Site-wide schema rollouts. Schema on pages with nothing quotable changes nothing
Publishing volume on learning-stage questions. Restart this once the buying prompts are covered

First ninety days

What we actually run. In the order it has to happen.

AI citation work has a real lag: crawlers have to refetch, indexes have to refresh, and answers vary between runs. The sequence below front-loads the things that move first.

The first ninety days. Nothing in block three starts before block one is answered.

Days 1 to 30: Find the real prompts

We assemble the prompt set from your sales calls, search data and the questions your team answers every week, then baseline which sources each engine cites today for each one. In parallel we check the unglamorous things: whether the assistant crawlers can reach you, and whether your company facts are consistent everywhere they appear.

Days 31 to 60: Write the citable pages

The content block. Each page takes one buying prompt and answers it in the first sentence, then supports it with sections that make sense in isolation, because that is the unit a model retrieves. Where a page already ranks, we usually restructure rather than rewrite, since the hard-won authority is worth keeping.

Days 61 to 90: Corroborate and measure

Corroboration is the slowest and most durable part: review platforms, partner directories, event agendas, podcasts and practitioner posts stating the same facts. Meanwhile we report citation share across the prompt set, crawler activity from the logs as the early signal, and any meetings arriving from assistant referrals.

What to measure

The numbers we report and the ones we refuse to lead with.

Metric	Why it is the right one here	When it should move
Citation share across the prompt set	The direct measure. Tracked per engine, because patterns differ and work for one usually lifts the others.	Early movement in weeks four to twelve, as pages are refetched
Presence in vendor-recommendation prompts	Being mentioned in an explainer is not the same as being named in a shortlist. These prompts are the commercial ones.	Slowest to move, usually one to two quarters
Assistant crawler activity on new pages	The earliest honest signal that pages are being seen at all, before citations change.	Two to four weeks after publishing
Sessions and meetings from AI assistant referrals	Low volume by nature and high intent. Judge it on meetings, never on sessions.	From the second month
Branded and comparison search volume	A buyer who met you in an answer often searches your name next. This catches the demand an answer created without a click.	From the second quarter

Reported, never led with: total prompts tracked; mentions without a link; schema coverage percentage; number of pages published.

Proof

Teams that arrived here. And what we built with them.

90 to 95%

of results from inbound

AI contract lifecycle management on ServiceNow. Category content and search took qualified meetings from single digits to tens a month.

Read story

15x+

return on every event attended

A Dubai crypto infrastructure company that could not run US ads. Search, events and ABM built its pipeline against Fireblocks-class rivals.

Read story

600K

monthly users, up from 50K in 11 months

A full website rebuild and a large content programme took HiddenBrains from 50,000 to 600,000 monthly users, through content marketing alone.

Read story

“What differentiated Lemniscate Growth is that they started with the buyer. On the inbound side, gated assets built for our category, a maturity assessment tool, and a search strategy that covers SEO alongside answer engine and generative engine visibility. On the outbound side, tiered account lists and multi-touch sequencing. We went from single digit qualified meetings in a month to tens of qualified meetings in a month. They operate like an extension of our own team.”

Sunil MasandHead of Product and Marketing, Aavenir

Why AI visibility is a middle of funnel job

It is tempting to file this under awareness, because it looks like visibility. The prompts that matter say otherwise. Buyers do not ask an assistant to introduce them to a concept and then buy something. They ask it to compare named options, to list alternatives to a tool they already pay for, to explain how a pricing model works, to say whether something integrates with their stack. Those are evaluation questions, which makes AI citation a middle of funnel channel wearing top of funnel clothes.

That reframing changes the work. You stop writing introductory explainers and start writing the material a buyer in evaluation needs, which happens to be the same material a committee needs and the same material a model can quote. One body of work, three payoffs. It is the clearest example of the 80:20 argument on this site.

The commercial prompts are comparison, alternatives, pricing and integration
Those are evaluation questions, so the content is middle of funnel content
The same pages serve buyers, committees and the models composing answers

What a limitation page does for you

A page that states plainly who should not buy your product is the highest-leverage page most B2B companies never publish. For a buyer, it is the fastest credibility test you can pass. For a champion, it is ammunition in a committee that suspects they are being sold to. For a model composing an answer, it is an unhedged, specific, attributable claim, which is exactly what gets quoted.

The objection is always that it will cost deals. In practice it disqualifies deals you were going to lose later and slower, and it shortens discovery on the rest because the buyer arrives having already read your honest version of the trade-offs.

Why rankings and citations are not the same thing

Ranking is a domain-level competition for a position on a page a human scans. Citation is a passage-level competition for a slot in the small set of sources a model was given. The two overlap, because retrieval for AI answers often starts from conventional search indexes, so a page that cannot rank frequently cannot be retrieved at all.

Where they diverge is authority. A mid-authority site with a definitive, well-structured, corroborated answer regularly gets cited over a larger competitor with vague content, because the model is choosing a passage rather than a brand. That is the opening for challengers, and it closes as categories mature, which is the argument for doing this work now rather than next year.

What this combines

AEO, GEO & SEO CXO branding Conversion optimization

Related situations

Traffic without pipeline Deals stall after the demo New market, no local proof

Questions buyers ask us. Answered plainly.

Still unsure? Ask us directly.

Is AEO different from SEO, or is this the same work with a new name?

It overlaps heavily and extends rather than replaces. Retrieval for AI answers often starts from conventional search indexes, so a page that cannot rank usually cannot be retrieved. What AEO adds is a set of requirements classic ranking work never enforced: self-contained passages, a direct claim in the first sentence, consistent entity data and third-party corroboration of your claims. We run them as one program with one report rather than two budgets.

How long before citations change?

Plan on sixty to a hundred and twenty days before drawing conclusions, and watch earlier indicators in the meantime. Assistant crawlers typically revisit new pages within two to four weeks, which is the first honest signal. Citation share on informational prompts moves next. Vendor-recommendation prompts, the commercially valuable ones, are slowest and usually take one to two quarters, because they depend on corroboration outside your own domain.

Can we win if our domain authority is lower than our competitors'?

More often than in classic SEO, yes. Models cite passages rather than domains, so a focused set of definitive, well-structured, corroborated answers can beat a larger competitor's vague pages. The strategy for a challenger is to pick narrow prompt clusters where you hold genuine expertise, answer them better than anyone, and get the same claims echoed in credible third-party places.

Should we block AI crawlers to protect our content?

That is a business decision with a clear trade-off, and we will give you the honest version. Blocking the assistant crawlers removes you from the answers they compose, which in most B2B categories means removing yourself from an early shortlist. If your content is the product, for example a paid research business, blocking may be right. If your content exists to create pipeline, blocking is self-harm. We check and document your current robots posture in the first thirty days, because teams are often blocking by accident.

Which engines should we prioritize?

Prioritize by where your buyers actually research, which for most B2B categories means the large general assistants and Google's AI features first, then the research-heavy tools. Track all of them in one panel, because citation patterns differ by engine while the work that earns the citation is largely shared. We report per engine so you can see which one your category concentrates in, rather than assuming.

Sources for the numbers on this page (2)

Google Search Central, AI features and your website, checked 4 October 2026. Query fan-out, how answers are composed from retrieved sources, and the available preview controls
Gartner press release, survey of 645 B2B buyers run August to September 2025, 20 May 2026. 45 percent of surveyed buyers used generative AI during a recent purchase, and 69 percent validate what it told them with a sales rep

Tell us where it is stuck. Twenty minutes is enough to find out.

Bring the revenue target, the account list if you have one, and last quarter’s pipeline. We will tell you which fifth of the work we would run first, and what we would stop.

20 minutes with a senior operator, not an SDR
Bring your revenue target and markets; we bring the pipeline math
Slots across US, Canada, India, Singapore and GCC time zones

Prefer email? growth@lemniscategrowth.com

Pick a 20-minute slotStraight to a senior operator. No SDR screen.