How do we show up in ChatGPT specifically?

Start with access, because it is the most common blocker. OpenAI runs three agents: OAI-SearchBot builds the index ChatGPT search draws on, ChatGPT-User fetches a page when a shopper asks about it, and GPTBot collects training data. Blocking GPTBot does not remove you from ChatGPT search; blocking OAI-SearchBot does, and a great many stores caught all three in one rule written in 2023. After access it is the same work as everywhere else: render on the server, state the facts as text, make the markup agree with the page.

Does blocking Google-Extended keep us out of AI Overviews?

No. AI Overviews and AI Mode are built from Google Search’s index, which Googlebot crawls, so a page that ranks can be summarised there. Google-Extended governs whether your content is used to train and ground Gemini models and apps. It is a training decision rather than a search one, and the two are set separately. We explain the trade-off before you choose either way.

ChatGPT says something about us that is wrong. Can you fix it?

Usually, but not by editing the model. Models restate what their sources say, so the fix is to correct the sources and wait for the recrawl: your own pages first, since they are the ones you control, then the profiles, listings, and marketplace entries that still carry the old fact. We log the description each engine gives you on every run, so you can see whether it has actually changed rather than checking by hand.

Why are we visible in one engine and not another?

Because they do not share an index, a crawler, or a retrieval method. Perplexity may have crawled you happily while OAI-SearchBot has been blocked since 2023, or Claude may be fetching a page that renders in the browser and getting nothing. That is why the audit and the report are both per engine: a single overall visibility number hides exactly the thing you need to act on.

A prompt set built from the questions your buyers actually ask, run across ChatGPT, Perplexity, Gemini, and Claude on a fixed cadence, with every cited domain logged, your position in the source list recorded, and the wording each model uses about you kept. The same questions every run, because a trend measured against a changing set of questions is not a trend. The first baseline is free.

Will good rankings get us into the answers?

They help with Google’s AI surfaces and Copilot, because those draw on search indexes you already compete in. They do less for ChatGPT search, which maintains its own index, and for Claude, which fetches during the conversation. Ranking is a useful head start and not a substitute, which is why the programme treats access and readability as their own work rather than assuming SEO covers it.

Can you guarantee a citation?

No. The engines decide what they quote, they change retrieval without notice, and the same prompt can answer differently twice in one day. What we guarantee is that the causes in your control are fixed, that the measurement is honest, and that the report shows the trend whether or not it flatters us.

Which platforms do you work on?

Shopify and Shopify Plus, Adobe Commerce, BigCommerce, Salesforce Commerce Cloud, Shopware, and headless storefronts. The engineers doing this work build on those platforms, which is why a rendering or schema fix here is a pull request rather than a recommendation you have to staff.

Scoped after the free scan and a thirty-minute call, because the work depends on what the scan finds. It runs month to month like every marketing engagement here, and both the scan and the first citation baseline are free whether or not you hire us.

Services
Digital Marketing
LLM and ChatGPT Visibility

See what
ChatGPT says.

Q: What is LLM SEO?

Making a store readable, quotable, and correctly described by large language models, so that when a shopper asks ChatGPT, Perplexity, Claude, or Google’s AI what to buy, your store is among the sources the answer is built from. It overlaps heavily with SEO on the technical side and differs in three places: access is set per engine rather than per search engine, content is judged on whether a fact can be lifted in one sentence, and success is measured in citations rather than rankings.

Be the answer. Not a footnote. See how ChatGPT, Perplexity and Claude describe you today, and what it takes to change it. Start with the free scan.

Run the free scan See our work

View case studies

Four engines, four answers

One question.
Four different answers.

The engines do not share an index, a crawler, or a taste in sources. A store can be the first citation in Perplexity and absent from ChatGPT on the same question. Six reasons the answers diverge.

Each engine runs its own crawler, and you can block one by accident
OpenAI’s OAI-SearchBot, PerplexityBot, and Anthropic’s ClaudeBot are separate agents with separate rules. A robots.txt written years ago commonly lets one through and stops another, and nothing reports it.
ChatGPT and Google reach you by different routes
AI Overviews and AI Mode are built from Google Search’s index, which Googlebot crawls, so a page that ranks can be summarised there. ChatGPT search draws on its own index. Ranking in one has no bearing on being read by the other.
Perplexity shows its sources, so the gap is visible
Answers carry a visible source list, which makes it the fastest place to see who is being cited for your category. If a reseller is sitting in that list and you are not, the reason is usually readability rather than authority.
Claude fetches on demand, and takes what it can parse
Answers are assembled from pages retrieved during the conversation. A storefront whose catalog appears only after JavaScript runs hands it a shell, and the answer gets built from whoever wrote plain HTML.
The description is wrong and nobody is watching
Models restate what the web says about a brand. Discontinued products, an old price, a positioning that has not been true for two years: all of it persists until the sources that say it are corrected.
None of it appears in any report you already read
There is no impressions chart for ChatGPT and no rank tracker for Claude. Without a prompt set run on a schedule, LLM visibility is managed on anecdote, usually whichever answer someone screenshotted last.

LLM and ChatGPT visibility

Engine
by engine.

The shared foundation is the same for all of them: reachable, server-rendered, factual, consistent. What differs is how each engine finds you and what it rewards, so the work is set per engine and checked per engine.

01 · ChatGPT

Three agents, three jobs

OpenAI runs OAI-SearchBot for the search index, ChatGPT-User for fetches during a conversation, and GPTBot for training data. Most stores that are invisible here blocked all three with one rule and meant to block only the last.

Per-agent robots rules, decided deliberately
Server-rendered product and category pages
Agentic checkout readiness

02 · Perplexity

Sources on display

Perplexity shows the sources behind an answer, which makes it the clearest read on where you stand and the quickest to respond when a page becomes quotable.

PerplexityBot access and index health
Answer-first openings on commercial pages
Comparison and alternative pages

03 · Claude

Read during the conversation

Claude retrieves pages as the conversation needs them, so what matters most is that the page a link points at is complete and parseable the moment it is fetched.

ClaudeBot access
Facts in text rather than in tabs or images
Policies and reviews as readable content

04 · Google and Copilot

The index you already rank in

AI Overviews and AI Mode summarise from Google’s index, and Copilot draws on Bing’s. Both reward the SEO you already do, plus the structured data that makes a fact safe to restate.

Googlebot and Google-Extended set separately
Entity markup and one consistent name
Bing index health for Copilot

05 · Measurement

What each engine says, weekly

One prompt set, four engines, the same questions every run. You see citation share per engine, who is cited instead, and how each model describes the brand in its own words.

Prompt set from your buyers’ questions
Citation share and source position per engine
AI referral revenue split out in GA4
Sentiment and factual errors logged

What goes wrong

Visible in one.
Absent in three.

Because the engines differ, the failures differ, and a fix that works for one can leave the others exactly where they were. Four we see repeatedly.

One robots rule, three engines lost
What it costs you
A line added to keep content out of model training also removed the store from the index ChatGPT search reads, because the agents doing those two jobs are different and the rule did not distinguish them.
What we do
Access decided per agent and written down: which crawlers index, which train, which you are refusing, with the trade-off stated before sign-off.
The model describes a business you no longer run
What it costs you
An old price, a discontinued line, a positioning from two rebrands ago. It repeats because the sources still say it, and every answer that repeats it costs a sale you never saw.
What we do
Correct the sources the model leans on: your own pages first, then the profiles and listings that disagree, then measure the description until it changes.
Cited in Perplexity, invisible in ChatGPT
What it costs you
The team sees a good answer in the engine they happen to use and assumes the channel is handled, while the engine with the larger audience has never successfully read the catalog.
What we do
The same prompt set run across all four engines, so the report shows the engine you are losing rather than the one you checked.
A screenshot instead of a measurement
What it costs you
Someone pastes a good answer into Slack and it becomes the status of the channel. The next week the same prompt answers differently and nobody notices, because nothing was recorded.
What we do
A fixed prompt set on a schedule with the raw answers kept, so a change in citation share can be traced to the answer that changed.

Free tool

Scan first.
Then decide.

Paste your store’s address. In a few seconds you see which engines can reach it, whether the catalog is readable without JavaScript, what your structured data says about you, and what to fix first. No email needed.

1. Crawler access, engine by engine
ChatGPT, Claude, Perplexity, Google AI, Copilot, Apple, and Meta, each with the verdict robots.txt gives it.
2. Readable without JavaScript
How much of the page a crawler that does not run scripts can read, and whether the product facts are in it.
3. Entity and product schema
Organization, WebSite, Product, reviews, and policies, and whether the markup agrees with the page.
4. What to fix first
Three findings, ordered by how much each moves the score, each with its measurement.

Takes: About ten seconds
Costs: Nothing. No email.
Reads: Public pages only

Run the free scan

The scan reads what any visitor’s browser receives. It does not run prompts against the engines; the citation baseline after the scan does.

Other options

A monitoring tool, a PR firm, or us.

An honest read on the alternatives. Each is good at something here.

Comparison criteria	RMG	AI monitoring tools	PR or reputation firm	In-house
Tells you what the models say	Four engines on a fixed prompt set, raw answers kept	Yes, and usually the reason to buy one	Sometimes, manually	Whoever remembers to check
Fixes why they say it	Crawler access, rendering, schema, and content, shipped as code	Not covered	Coverage and profiles, not the storefront	Roadmap permitting
Corrects third-party sources	Profiles, listings, and marketplace facts brought into agreement	Not covered	This is the thing they are good at	Rarely anyone’s job
Ecommerce depth	Catalog, feed, and checkout readiness on your platform	None; it is a tool	Rarely	Your own stack, and only that
Commitment	Month to month	Monthly subscription	Retainers, often annual	Headcount

Results

Show you the money.

DTC · SEO + CRO
+202%
traffic · month over month
A full SEO and CRO rebuild. Organic search now drives 25% of overall revenue, and AOV rose 30%.
DTC · Shopify Plus
2.4x
faster page loads · zero SEO loss
Replatformed an aging custom site to Shopify Plus without losing a single SKU or SEO ranking.

See all case studies

The big numbers

+30%

Avg. conversion rate lift within 12 months

+50%

Avg. order value lift

<1s

Page load speeds on the storefronts we build

+15%

Avg. revenue lift within 6 months

How it starts

Here's what happens
after you fill out the form.

Sales keeps it short. A senior strategist and an engineer who has shipped on your platform join the first call when the question needs them.

Step 01
30-min
discovery call
What you’re on, what hurts, what the ERP and the catalog look like. We’ll say early if the platform isn’t the problem.
Step 02
Free platform audit
Pricing logic, catalog, integrations, Core Web Vitals on real field data. Five business days. Yours to keep.
Step 03
Stay, upgrade,
or move
One recommendation, with the cost of staying set against the cost of moving. Sometimes the answer is stay.
Step 04
Scoped
proposal
Scope, timeline, pricing: fixed-fee or T&M. First wins inside 30 days.

WE GOT THE WHOLE PACKAGE

There’s a lot more
to this story.

Design, development, and digital marketing all live under one roof. The same team that builds your store also designs it and runs the paid, lifecycle, and CRO work behind it, so nothing gets lost in a handoff.

View services

THE QUESTIONS YOU HAVE -

Get them out
of the way.

Making a store readable, quotable, and correctly described by large language models, so that when a shopper asks ChatGPT, Perplexity, Claude, or Google’s AI what to buy, your store is among the sources the answer is built from. It overlaps heavily with SEO on the technical side and differs in three places: access is set per engine rather than per search engine, content is judged on whether a fact can be lifted in one sentence, and success is measured in citations rather than rankings.

Also on the menu

One team.
The rest of the work.

Each of these is its own page because people search for it by name. Behind the pages it is one team, so the SEO work knows what the developers shipped and the email flows know what the ads promised.

All of Digital Marketing Every service

Free AI visibility scan

What do they
say about you?

Run the scan first; it takes seconds and asks for nothing. Then say what you want done about it, and a senior strategist replies within one business day with a citation baseline across ChatGPT, Perplexity, Gemini, and Claude.

Already know what you want fixed? Say so here and we will scope that instead.

See what
ChatGPT says.

One question.
Four different answers.

Each engine runs its own crawler, and you can block one by accident

ChatGPT and Google reach you by different routes

Perplexity shows its sources, so the gap is visible

Claude fetches on demand, and takes what it can parse

The description is wrong and nobody is watching

None of it appears in any report you already read