Home  /  Services  /  AI Search Visibility

08 · AI Search Visibility · Fixed scope

Your buyers now ask ChatGPT before they search. The answer names your competitor.

"Which provider is right for a company like ours?" ChatGPT, Gemini, Claude and Google's AI Overviews answer that question by picking three names. Three names instead of ten links. You can sit on page one of the search results and still be missing from those three — in which case the conversation ends without you, and the click never happens. Most companies have never checked whether their name is in there.

This project measures that first, then fixes the reasons you are not cited. It takes 25 questions buyers actually ask and records what four engines answer, makes your entity something an AI can be certain about, opens the site to AI crawlers, rebuilds your five most important pages into a form that can be quoted, and gets you listed on the sources AI answers rely on. Four weeks later it asks the same questions again and shows you the before and after as it is.

Who this is for

Three situations we see most.

B2B · SaaS

Buyers ask an AI for "recommended providers", and only competitors come up.

You rank third on Google, but ask ChatGPT the same question and you are absent while two competitors and a review site are named. The shortlist is built inside the conversation, and sales only reaches companies on that list.

E-commerce · DTC

Product recommendations are decided on review sites and Reddit.

The sources behind "a good product for X" are someone else's review pages, not your site. If those sources do not cover you, the AI does not know you exist.

Any company

Rankings held. Clicks fell.

When an AI overview shows the answer first, users click links about half as often. Holding a position is no longer enough; you have to be the source cited inside that answer.

Not for you ifYou are a well-known brand whose traffic is mostly brand search, or a new domain without ten pages worth citing yet. Both are in the honest limits below, and the free website audit will say so.

Why it matters

AI does not rank. It cites.

Ten links become three names

Search showed candidates and let the user choose. An AI answer has already chosen. The bar for being a candidate moved from "page one" to "one of three names in the answer".

An AI cites you only when it is sure who you are

If your name is written differently on every page, there is no schema, and nothing outside your site describes you the same way, you are an unverified entity. Sources an engine is unsure about do not get cited.

What cannot be read cannot be cited

If robots.txt blocks AI crawlers, the body only renders in JavaScript, or the answer is buried at the bottom of the page, even excellent content never becomes material for an answer.

Almost nobody measures it

AI citations do not appear in Search Console the way rankings do. The only way to know is to ask and record, so most companies have no idea where they stand. What is not known cannot be fixed.

What you get

Eight parts, all on your own site.

01 · Baseline measurement

We agree 25 questions buyers actually ask, then ask each in Google AI Overviews, ChatGPT, Gemini and Claude and record the answers: whether you were cited, who was cited instead, and which sources the answer relied on. The starting point is documented.

02 · Entity consistency

Company name, description, address, founding date and services made identical to the letter across the website, Google Business Profile, LinkedIn and major directories, and linked with sameAs in Organization and Service schema, so an AI can be certain it is one company.

03 · Crawler access and llms.txt

robots.txt reviewed for GPTBot, ClaudeBot, PerplexityBot and Google-Extended, a check that the body renders without JavaScript, an llms.txt site summary written, and Bing Webmaster Tools connected with IndexNow. ChatGPT search draws on the Bing index.

04 · Structured data

FAQPage, Article, Service and Organization schema fitted to each page type. What exists is corrected, what is missing is added, and all of it passes validation.

05 · Five key pages restructured

Your five most important pages rebuilt answer-first: question headings, the direct answer in the first paragraph, definitions and comparison tables, figures with sources. Citations do not come from "good writing"; they come from paragraphs that can be lifted.

06 · Citation sources

Among the external sources the baseline shows AI answers relying on — review sites, directories, communities, wiki-style pages — we pick five where you can be listed and create or correct the profile. An AI trusts places where others describe you.

07 · Re-measurement and report

In week four the same 25 questions go to the same engines and are recorded again: questions now cited, pages newly cited, questions still not cited and why. We do not show only what improved.

08 · 90-day playbook

The pages to write answer-first next quarter, a measurement sheet and procedure that takes thirty minutes a fortnight, and a checklist for adding pages without breaking schema. So your team can carry on.

What lands in the folderThe 25 questions with the baseline and re-measurement sheets, the entity definition, the schema applied with validation results, robots.txt and llms.txt, the five restructured pages live on the site, the list of sources secured, the before-and-after report and the 90-day playbook. Nothing sits in an account of ours.

How it is built

Measure first, fix, measure again.

01

The questions are the buyer's, not ours

The 25 questions come from sales calls, enquiry emails and support logs, not from a keyword tool. The measurement only means something if these are sentences people actually type into an AI. The list is approved by you and does not change for four weeks.

  • Four engines — Google AI Overviews, ChatGPT, Gemini and Claude, in order of usage, each with its own citation behaviour. Perplexity is added for technical and B2B clients who want it.
  • Same conditions — logged out, fresh session, same country setting, so before and after are comparable.
  • Competitors recorded too — whoever fills the space you are missing from is the to-do list.
02

Make yours a company an AI can be certain about

A citation is the result of trust. An engine picks companies whose facts it can confirm across several sources. That is why aligning the schema inside the site with the profiles outside it, under one definition, comes before any content work.

  • One-sentence definition — who, for whom, what — the same sentence everywhere.
  • Linked with sameAs — the site, LinkedIn, Google Business Profile and directories point at each other.
  • Our own site is the evidence — growlinq.co runs llms.txt and schema the same way.
03

Written in paragraphs that can be lifted

An AI does not cite a whole page. It lifts the two or three sentences that answer the question directly. So restructuring is about structure, not style: under a question heading, the first sentence is the answer and what follows is the evidence.

  • Answer first — directly under the heading, not after three paragraphs of preamble.
  • Comparison tables and definitions — the formats AI engines lift verbatim most often.
  • Figures with sources — unsourced numbers are not cited, or are cited and cost you trust.
04

Measured honestly

AI answers vary slightly every time, even to the same question. So we do not present "a 40% citation rate" as a precise figure. We measure twice under the same conditions to see direction, and the report says in plain words that this is a sample.

  • Sample-based — 25 questions, once per engine, conditions fixed. An observation, not a statistic.
  • Cross-checked in GA4 — sessions referred from chatgpt.com, gemini.google.com, claude.ai and the like are grouped as their own channel to show real visits.
  • What did not work is written down — questions still not cited, and why, are half the report.
05

Korea, if it is in scope

Korean users also ask Naver AI Briefing and CLOVA X. Naver preferentially cites its own Blog, Cafe and KnowledgeiN, so a Korean question set is measured separately and citation sources are found inside Naver.

  • Ten Korean questions added — written the way Korean buyers phrase them, not translated.
  • Naver sources — Blog, Cafe and Smart Place become the citation sources. Scoped together with our Korea market entry work.

Process

Four steps, about four weeks.

Week 1

Questions and baseline

25 questions agreed, baseline taken across four engines, competitors and sources recorded. Entity definition written and approved.

Week 2

Foundations

robots.txt, llms.txt and Bing connected; schema applied and validated; company facts on external profiles brought in line with the definition.

Week 3

Content

Five key pages restructured answer-first and published on the site. Profiles on five citation sources created or corrections requested.

Week 4

Re-measure and hand over

The same 25 questions measured again, the before-and-after report, AI referral channel set up in GA4, the 90-day playbook and a working session.

What we measureThe change in how many of the 25 questions cite you, before and after, and sessions arriving in GA4 from AI engine referrers. Both are samples that show direction, and the report says so. Rankings and total traffic are not reported.

Honest limits

What this project will not do.

01

It does not guarantee citations. AI engines publish no ranking table and give different answers to the same question. What we do is put every condition for being cited in place, and show the re-measurement as it is, good or bad.

02

It does not create authority you do not have. An AI trusts the places where others describe you. A company with no reviews, no press mentions and no customer stories is unlikely to be cited within four weeks. In that case we tell you what to do before this project.

03

The measurement is a sample. 25 questions, asked and recorded by a person. It is not the thousands of prompts and daily tracking of a paid monitoring tool, and if you reach the scale where that is needed, we will say so.

04

It does not write more than five pages. This project builds the foundation and the method; adding new answer-first pages every month is the job of the SEO & Content retainer. The playbook is the handover document between the two.

05

If you are already being cited, you do not need it. If you appear in more than half of the 25 questions, the foundation is already there and what remains is content velocity. The free website audit asks a few of the questions first and tells you so.

FAQ

AI search visibility, answered.

How is this different from SEO?+
The foundations overlap, and AI search visibility is in fact one line item inside our SEO & Content retainer. The goal differs. SEO targets a position in a list of links; this project targets being cited inside an AI-generated answer. So it weights entity certainty, crawler access, liftable paragraph structure and external sources more than rankings. This project builds that foundation once, in four weeks; the retainer keeps adding content on top of it.
Which AI engines do you measure?+
Google AI Overviews, ChatGPT including its search mode, Gemini and Claude — in order of real-world usage. AI Overviews sit on top of Google results, so their reach is the largest. Perplexity has a small user base concentrated among technical and B2B buyers, so for those clients it is added as a fifth engine.
What do you need from us to start?+
Edit access to the site's CMS or a developer contact, management rights on Google Business Profile and the LinkedIn company page, GA4 edit access, and one person who can help define the 25 buyer questions — ideally from sales or customer support. If the questions are not the ones buyers actually ask, the whole measurement is meaningless.
How long does it take?+
About four weeks from access: questions and baseline in week one, foundations in week two, content and sources in week three, re-measurement and handover in week four. Approval of profiles on external sources follows their schedules, so some may land after handover, and the report says so.
How quickly do results show?+
It depends on the engine. ChatGPT search follows the Bing index and tends to reflect new pages within days to weeks; Google AI Overviews and Gemini move with the Google index. Claude reads the live web when search is on, so it reflects changes quickly but its answers vary more. The week-four re-measurement shows early direction; the playbook includes repeating the same sheet at ninety days.
Should we not be blocking AI crawlers?+
Training crawlers and search-and-answer crawlers are different, and most can be controlled separately. To be cited, the answer crawlers must be allowed. Whether to block training crawlers is a content policy question you decide; we explain the difference and the consequences of each choice, then configure it.
Does this cover Korea?+
When Korea is in scope, ten Korean questions are added and measured separately in Naver AI Briefing and CLOVA X. Naver preferentially cites its own Blog, Cafe, KnowledgeiN and Smart Place properties, so citation sources are secured inside Naver. This is usually scoped together with our Korea market entry work.

Also

SEO & Content →The retainer that keeps adding answer-first pages on the foundation built here. Measurement & Tracking →To capture AI referrals as a channel in GA4, measurement has to be accurate first. Korea Market Entry →Naver AI Briefing, CLOVA X, and the citation sources inside Naver.

If you have read this far, the website audit is free.

The free audit is a website audit: how you show up in search and in AI search, whether basic tracking is alive, and the first three things we would fix, in writing. It is the same audit whichever page you came from, and it costs nothing. If it is AI Search Visibility itself you want, say so in the same form. The audit asks a few of the questions first, and we answer whether you need this project at all.

Get a free website audit

Or ask about AI Search Visibility →