Free AI Visibility Checker

What this tool checks — and what no tool can check

Type your website in and this tool reads it the way a crawler does: your homepage, your robots file, your sitemap and a sample of up to ten pages. It then scores eleven categories of signal — the things an AI assistant's sources can actually see about your practice — and returns a number out of 100 with the working shown.

It is worth being blunt about the other half. Nobody can tell you where you rank inside ChatGPT, because there is no ranking to read. Assistants do not publish a results page, their answers change between users and between runs, and asking one about yourself tells you almost nothing — models are agreeable, and a model shown your name will often say something flattering about it. Any tool claiming to measure your ChatGPT position is either guessing or asking the model and reporting the guess back to you as data.

So this tool measures the part that is real and that you control. When an assistant answers a question about interior designers in your city, it is working from sources: web pages it has been trained on, search results it retrieves at the moment of asking, directories, review platforms and press. Every one of those is downstream of what exists about you in public. That is what we score.

How an assistant actually arrives at a name

When somebody asks an assistant to suggest an interior designer in Noida or an architect in Pune, roughly three things happen, and each one has a failure mode you can do something about.

1. 1It works out what is being asked for. A service, a city, a budget band, a property type. If your website never states plainly which of these you do, you are not a candidate for any of them.
2. 2It gathers sources. Depending on the product, that means retrieving live search results, reading indexed pages, or drawing on what was in its training data. All three are text. A gallery of photographs with no writing around them contributes nothing at this stage.
3. 3It writes an answer, usually naming a handful of businesses. It favours things it can state confidently — a practice whose name, city, services and body of work are consistently described across several independent sources is far easier to name than one that appears once, ambiguously.

The practical consequence is that this is not a new discipline so much as an unforgiving version of an old one. The work that makes you findable in AI answers — being a clearly-defined entity, writing properly about what you do, being described by other people — is the same work that makes you findable in search. What has changed is the tolerance. A search engine will list you tenth. An assistant names three or four firms and stops.

The eleven signals this tool scores

Each category is scored out of ten. Eight are read directly from your website. Three cannot be — they live on Google, on other people's sites and on platforms a crawler is not permitted to read — so the tool asks you about them instead, and leaves them out of your score entirely if you would rather not answer.

How the score is calculated

Every category holds a handful of individual checks, each worth a set number of points. A check either passes, partly passes, or fails — and each one shows you what we saw, so you can disagree with it. The category score is the points earned as a share of the points available, out of ten. Your headline score is the average of your scored categories, multiplied by ten.

The important part is what happens to a category we cannot assess. It is withheld, not failed. It shows as a dash, it is excluded from the average, and your score is calculated across the categories that had real data. A tool that scored an unasked question as zero would be reporting on its own gaps and calling it your website's problem.

A high score does not mean an assistant will name you tomorrow, and a low one does not mean it never will. It means your website is or is not carrying the signals that make naming you possible. That is a real thing to know, and it is the honest limit of what a website audit can tell you.

Letting the crawlers in, before anything else matters

Every other category on this page assumes something that is often not true: that the crawlers feeding these assistants are allowed to read your website. A single line in a file called robots.txt can withdraw that permission, and it is the one failure no amount of good content compensates for. You can have the schema, the depth, the city pages and the reviews, and still be a closed door.

It is also the failure that hides best. It changes nothing in a browser, it does not show up in Search Console alongside your Google coverage, and it usually does not touch the Google rankings most practices watch. A developer tidying up a robots file, or a plugin writing one for you, can shut out an assistant's crawler on a Tuesday afternoon and nobody finds out.

The tool parses your robots.txt the way the standard says it should be read — per user-agent group, longest match wins, the wildcard only as a fallback — and then tests every page it actually fetched against those rules. It names the crawlers individually, because they are not one thing:

• Answer-time crawlers fetch a page at the moment somebody asks a question. OpenAI's OAI-SearchBot and ChatGPT-User, Anthropic's Claude-User and Claude-SearchBot, PerplexityBot and Perplexity-User, and Google-Extended. Block one of these and you are out of answers being written this week.
• Training crawlers gather text that may inform a later model — GPTBot, ClaudeBot, CCBot, Applebot-Extended, meta-externalagent. Blocking one costs you nothing today and compounds quietly for years.
• JavaScript-only pages. Several of these crawlers read the HTML your server sends and nothing more. If that HTML is an empty shell that fills itself in after loading, there is nothing to read, and the tool says so.

Being a clear entity, not a nice-looking site

The single most common failure we see on design and architecture websites is beautiful and anonymous. Enormous hero images, a one-word headline, a scroll of photographs, a contact form. It works on a human, who can infer from the photographs that this is a design or architecture practice. A machine reading the same page finds a title that says Home, no description, no About page worth the name and no structured data. It cannot establish that you are a business, what kind, or where.

Fixing this is unglamorous and fast. Your homepage title should contain your practice's name and what you do. You need an About page that reads like a description of a company rather than a mood statement — when you started, who runs it, what you specialise in, where you work. Your social profiles should be linked from your site and declared in your markup, because that is how a machine confirms that the practice on Instagram and the practice on this domain are one entity.

Writing enough that there is something to quote

Retrieval systems work with passages. If a page has forty words on it, there is no passage — nothing can be pulled out of it and attributed to you. Interior design websites are unusually prone to this, because the work is visual and the instinct is to let the photographs speak.

The tool checks how many of the pages it read carry enough text to be useful, and how substantial your homepage is. It is not asking for padding. It is asking whether a page about turnkey interiors explains what turnkey means in your practice, what is included, how long it takes and what a client is signing up to — the things you already say on every call, written down once.

• Every service gets its own page. One page listing eight services makes you a candidate for none of them.
• Each of those pages explains the service, not just names it. Scope, process, what changes the price, what goes wrong.
• Every page gets its own title and description. Duplicates tell a crawler these pages are interchangeable.
• Write the specifics you would say out loud: the material, the timeline, the constraint you solved.

Naming your city, in the places that count

Almost every question that could name your practice has a place in it. Somebody asks for a designer in Gurgaon, an architect in Hyderabad, a turnkey contractor near Whitefield. If your city is mentioned once in your footer address and nowhere else, you have not connected your practice to the place it works in.

The checks here are specific: does your city appear in the homepage title, heading or description; does it appear across the site rather than on one page; is there a machine-readable address; is there a page for the area you actually serve. The last one matters most for practices that take work across several cities — a paragraph on your homepage saying you work nationally is much weaker than a page per place that says something true and different about each.

Proof that does not come from you

Three of the eleven categories are about evidence that exists outside your own website, and they are weighted heavily for a reason: a claim you make about yourself is worth very little to a system deciding whether to put your name in front of somebody.

Your project work, written up

A portfolio grid is not evidence. A project page is: this flat, this size, this brief, these constraints, this is what we did and this is what it cost in time. That page can be cited. A thumbnail cannot. The tool checks whether your projects have their own pages, how much is written on them, and whether your images carry alt text — which is the only thing a text-based reader sees of a photograph.

Reviews, and their absence from your own site

We do not read your Google reviews. Scraping Google's results is against their terms and we are not going to put a client's domain at risk to populate a free tool. So the tool asks you for your rating, your review count and how many arrived in the last ninety days — recency matters, because a practice with forty reviews from four years ago reads as dormant. It also checks something we can see: whether your reputation appears anywhere on your own website in a structured form, which is where most designers have done nothing at all.

Being written about by other people

Directories, press, a guest article, an award, a brand that lists you as a partner. Each is an independent source that says you exist and does not belong to you. This is the hardest category to move and the most durable once moved, and it is usually the lowest score on an otherwise well-built site.

Structured data: saying it in a format that needs no interpretation

Structured data is a block of JSON in your page source that states the facts about you plainly — this is an organisation, this is its name, this is its address, this is a service it offers, this is a question and this is its answer. Nothing is inferred. It is the difference between a machine deducing your phone number from a page and being handed it.

The tool looks for the types that matter for a design practice, and it parses them properly rather than pattern-matching, so markup that is present but broken will not score. In order of value:

• Organization or LocalBusiness — who you are, with your address and linked profiles.
• FAQPage — question-and-answer pairs, the single most quotable format on the web.
• Service — each service as a described thing, with the area you serve.
• BreadcrumbList — where each page sits, so a page is read in context.
• Article or BlogPosting — your written content, attributed and dated.
• Review or AggregateRating — your reputation, stated on your own domain.

What to fix first

Your result comes with a prioritised plan, but the order is usually the same, because the failure modes are. If you want to start before you run the audit:

1. 1Open yoursite.com/robots.txt and read it. Two minutes, and it is the only item here that can cancel out all the others. If any block names GPTBot, ClaudeBot, PerplexityBot, Google-Extended or the OpenAI and Claude answer-time crawlers, decide deliberately whether you meant that.
2. 2Make your homepage say what you are and where. Title, heading, description. This takes an afternoon and it is the single biggest unlock on most design websites.
3. 3Add Organization and LocalBusiness markup. One block of JSON, once, with your real details in it.
4. 4Split your services onto their own pages and write several hundred words on each about how you actually work.
5. 5Write up three projects properly. The brief, the constraints, the materials, the timeline, the outcome.
6. 6Publish the ten questions clients ask before they sign, marked up as an FAQPage. Cost drivers, timelines, what is included, what goes wrong.
7. 7Get listed and get written about. Directories first because they are quick, then one real piece of press or a guest article.

None of this is exotic. It is the work of describing your business clearly enough that a stranger — or a machine acting on a stranger's behalf — could repeat it accurately. If you would rather have someone do it with you, our SEO for interior designers service covers exactly this ground, and the Google Business Profile audit handles the half of your local presence that lives on Google rather than on your site.

What this score is not

Worth stating plainly, because the category is full of tools that will not.

• It is not a ChatGPT ranking. No such ranking is published, and we do not pretend to infer one.
• It is not a claim that any assistant recommends you today, or that it will after you fix things. We cannot see inside those systems and neither can anybody selling you a report that says otherwise.
• It is not an official metric from OpenAI, Google, Anthropic or Perplexity. It is our assessment.
• It is not a full site crawl. We read a sample of up to ten pages within a time limit, chosen to cover your homepage, services, projects, contact and writing. A large site will have pages we did not reach.
• It cannot see what your site renders in JavaScript after loading. Neither can many of the crawlers that feed these systems, which is why we flag it when we find it.
• It is not advice about anyone else's website. Run it on your own.

Related tools & reading