Zero-Click Visibility

The problem it solves

Ask an AI tool which vendors to look at and you get a short list, a sentence about each, and maybe a few links. Most people read the list and move on. The shortlist now forms inside the answer, before anyone visits a website.

Most teams still measure this the way they measure search: rankings and clicks. Neither one tells you what the answer said about you, whether it was true, or whether you were on the list at all.

Named beats cited

For a long time the goal was the citation, the little link under the answer. Citations still matter, because the pages engines cite shape what the answer says. But most buyers never open them. What they read is the list of names.

The goal is to be named, accurately, on the questions buyers ask right before they choose.

That means the bottom of the funnel. The questions worth winning look like these:

• "Best [category] platforms for [use case]"
• "[Competitor] alternatives"
• "[You] vs. [competitor]"
• "How much does [category] cost?"
• "Is [you] any good?" and "[You] reviews"
• "Which vendors can [specific requirement]?"

Top-of-funnel definitions and how-tos still earn citations, and they're worth having. They just don't decide who gets on the shortlist.

Three questions to answer first

The question	What it tells you	Fix it with
Is the information about us accurate?	Whether the engines get your basic facts right: what you do, who you serve, when you were founded, where you're based, who runs it, what you've been recognized for	Your own pages first, then the public record
Are we showing up?	How often you're named on unbranded, bottom-of-funnel questions, and where you land on the list	First-party evaluation pages, reviews, roundups, communities
Are we named, or only cited?	Answers that link to your page without ever saying your name. Your content is doing the work and someone else is getting the credit	Pages that state who you are in the sentence the engine is likely to lift

Answer them in that order. If the engines describe you wrong, more visibility only spreads the wrong description further.

Where answers come from

The engines build answers from far more than your website. These are the places to look, and later the places to work.

Place	What to check	Mostly earns
Your own site	Do your facts match on every page, including old press releases, PDFs and help docs?	Accuracy, citations
Encyclopedias and knowledge panels	Is the entry current? What category does it put you in?	Accuracy
Directories and registries	Are you listed where engines look when someone asks what tool to use?	Names
Review sites	How many reviews, how recent, and what do they say?	Names, trust
Third-party roundups and comparisons	Do the "best of" and "X vs. Y" pages engines cite include you, and get you right?	Names
Communities	What practitioners say about you on Reddit, forums and Slack groups, and in what words	Names
Social and video	LinkedIn posts, newsletters and YouTube transcripts, which engines read as sources	Names
Earned media and analysts	Whether coverage and analyst recognition actually show up when someone asks who leads the category	Names, trust
Developer channels	Public docs, code repositories, and the places developers ask for tool recommendations	Names, citations

What data to gather

Before you run anything

• An answer key. Your canonical facts, one row each, with the exact correct wording and any wording you never want used. Settle disagreements inside the company first.
• A prompt set in buyers' words. Branded questions to test accuracy, and unbranded bottom-of-funnel questions to test whether you show up.
• A competitor list. Who you expect on the shortlist, so you can measure share of answer.

For every answer

• Were you named? Where on the list?
• Were the facts right? Which ones were wrong?
• How were you described, and would you describe yourself that way?
• Which pages did it cite, and do those pages name you?
• Which competitors were named?

Alongside it

• Which domains and pages engines cite most for your category, yours and everyone else's.
• An offsetting metric you can trust: branded search, direct traffic, traffic from AI tools, and pipeline. AI answers aren't repeatable, so visibility on its own is a vanity metric.

Where the data comes from

Source	What it gives you
Sales calls, customer interviews, communities	How buyers actually phrase the question. This is Field Research, and it's where the prompt set should start.
Keyword research tools	Which bottom-of-funnel questions have real demand
The AI engines, queried directly	The answers themselves, with web search turned on so you see what buyers see. An API makes it cheap and repeatable.
Search and AI data services	Which domains and pages engines cite across thousands of answers, and where you appear in AI search results
Brand and social listening	Volume and sentiment of mentions, and the vocabulary practitioners use
Your own analytics and CRM	The offsetting metrics: branded search, direct traffic, AI referrals, pipeline

How to record a baseline

1. Lock the prompt set. A core set of about 40 questions for the monthly pulse, and a fuller set of about 150 for the quarterly audit. Change it only at quarter boundaries, or the trend means nothing.
2. Run every prompt across the engines your buyers use, with web search on. Run a slice of it twice to check stability.
3. Score each answer against the answer key: named, position, accurate, cited, competitors named. Automate the scoring, then have a person review it. Loose scoring makes errors look like wins.
4. Save the raw answers. When a rule changes, you rescore old runs instead of starting over.
5. Log the model versions. Treat a model change as a break in the trend line.
6. Write the readout: a grade per topic, the live fact errors with their sources, and where you're missing from the shortlist.

The sequence

When	What happens	Why then
Week 1	Answer key, prompt set, baseline run, readout	You can't show movement without a starting line
Month 1	Fix every fact on pages you control. Publish a facts page written for AI engines.	Cheapest fix, and every later step depends on it
Months 1–2	Correct the public record and get listed in the directories that matter	Engines lean on these for basic facts and "what should I use" questions
Months 2–3	Pilot first-party bottom-of-funnel pages: comparisons, alternatives, pricing approach. Start outreach and a review drive.	This is where the shortlist questions get answered
Months 3–6	Scale what the pulse says is working. Community, social and earned media on a steady clock. Original research if there's a story.	Being named by others takes months, so start early and keep at it

Each action runs Pilot, Signal, Scale-up: ship the smallest version, check the next monthly pulse, and build the full version only when the signal shows. Run an extra pulse two to four weeks after any big fix. It's Napkin Math applied to search.

Example changes

On your own site

• Make founding date, headquarters, leadership, ownership and recognition identical everywhere, old press releases and PDFs included
• Publish one page of company facts written for AI engines, and keep it current
• Put a plain sentence saying who you are and what you do near the top of the pages engines already cite
• Build comparison and alternatives pages for the matchups buyers actually ask about
• Say how your pricing works, so engines stop repeating someone else's guess
• Move product documentation out from behind a login
• Use the words practitioners use, even where they differ from your positioning deck

Off your site

• Get listed in the directories and registries engines check for your category
• Run a review drive on the review sites that show up in answers
• Ask roundups that already list you to update outdated details
• Pitch roundups that leave you out, with a real reason to add you
• Show up in communities as people, consistently, long before you need anything
• Publish original data that journalists and engines can attribute to you

Getting other people to change their pages

Start with the pages engines cite most often for your bottom-of-funnel questions. A wrong page nobody cites can wait.

The page	The ask
Lists you, with outdated details	A short, friendly update request with the exact corrected sentence and a source link. The easiest yes you'll get.
Leaves you out	An inclusion pitch with a reason that helps their readers: new data, a new capability, a perspective they're missing
Gets you wrong and won't fix it	One factual correction, sent once. Then publish a better first-party page on the same question and let it compete.
Encyclopedia entries	Request changes through the talk page with independent sources. Never edit your own entry.
Wants payment for placement	Only as a clearly labeled sponsored placement, or not at all

Log every ask: the URL, how often engines cite it, what you asked for, when, the response, and whether the engines picked up the change. The last column is the one that matters.

What success looks like

Four numbers, measured against the baseline every month:

Metric	What it is	Where it should go
Fact accuracy	Share of branded answers with the facts right	Up first, and fastest
Live fact errors	Specific wrong facts still circulating, each traced to a source	To zero
Named rate	Share of unbranded bottom-of-funnel answers that name you	Steadily up, topic by topic
Share of answer	How often you're named relative to the category leader	Closing the gap

Set three- and six-month targets with leadership once the baseline is in. Expect accuracy to move within a quarter and the named rate to take two or more. Models update slowly, so the first months mostly show whether your fixes are being picked up.

A few rules keep the numbers honest:

• Moves under about five points are noise unless they hold for two runs.
• Never add numbers from different tools together. Report them side by side.
• A flattering answer that's wrong still counts as wrong, and you don't amplify it.
• Every visibility number gets reported next to its offsetting metric.

Where it's running

I'm running this now at Level Access, across five AI answer engines. The baseline ran in October 2026, with checkpoints at the end of January and April 2027. The first version was a guide to AI search I built at Fleetio. Results will be posted here when they're in, good or bad.