How can we help?

Noma measures how AI assistants describe and recommend your brand. It asks five engines the same set of category questions every day, records which brands each answer names, which URLs it cited, and how the answer reads, then turns that into a visibility score you can track over time.

A project is one brand and one domain. Enter your website and Noma reads your public site to work out what you sell, who you compete with, and what buyers would ask. You can edit all of it afterwards; nothing is locked in by the wizard.

Competitors are set per project under Competitors. Noma proposes a starting set from what the engines already name alongside you, which is usually more accurate than a list drawn up internally, because it reflects who the models actually consider comparable.

Overview carries the three headline numbers: visibility, share of voice and sentiment. Everything else drills into one of them. Sources shows what the engines cited, Gap analysis shows the prompts you are missing from, and Health shows the technical audit of your own site.

Each tracked prompt is sent to every active engine once a day from a fresh session, so nothing carries over between runs. The answer is stored, then parsed for brand mentions, competitor mentions, cited URLs and sentiment.

Five: ChatGPT, Claude, Gemini, Grok, Perplexity. Each is reported separately rather than averaged, because being invisible on one engine and strong on four is a different problem from being weak everywhere, and a blended number hides it.

Visibility is the share of your tracked prompts where an answer named your brand, across all engines in the selected period. Forty prompts on five engines is two hundred answers; if your brand appears in twenty of them, visibility is 10%.

Your share of all brand mentions inside the same answers. Visibility can rise simply because a category got busier; share of voice only rises when you gained ground against the brands you appear alongside.

Usually because the engines genuinely do not name you yet, which is the finding rather than a fault. Check two things first: that your prompts are category questions a buyer would ask rather than branded ones, and that your brand's alternate spellings are registered, since a mention written differently will not match.

The models are not deterministic and the web they read changes daily. Single-day movement is usually noise; a trend across a week or more on a fixed prompt set is signal. This is why the prompt set matters more than any single number.

Two levers, in order. Earn mentions in the sources Noma shows are already cited for your category, because third-party pages carry more weight than your own claims. Then fix the signals on your own site that decide whether you can be parsed and trusted, which is what the 12-point audit under Health measures.

Yes, under the project's brand settings. Add every form the models might use, including the legal name, common misspellings and any product name used interchangeably with the brand. Mentions are matched against all of them.

Being cited and being named are different outcomes. An engine can quote your page as a source without mentioning your brand in the answer text. That is a content problem rather than an authority one: the page is trusted, but nothing in it makes the brand the answer.

Under Prompts. Add the questions a buyer would actually type when they do not yet know you exist. Prompts can be paused rather than deleted, which keeps their history intact for comparison.

Enough to cover the ways your category gets asked about, and then stop changing them. Comparability depends on the set staying fixed: adding prompts where you happen to be strong raises the number without anything changing in the world.

Tags group prompts so you can read visibility for one part of the business at a time, under Tags. A prompt can carry several. Filtering the dashboard by tag scopes every panel to that group.

A prompt earns its place if a real buyer would type it before they know your brand. Branded prompts ("is Acme any good") will nearly always mention you and flatter the total, so keep them separate from category prompts, which are the ones that decide a shortlist.

Yes. Every run stores the full response, so any number on the dashboard can be traced back to the exact text and the model that produced it. That matters most when an engine states something untrue: you have the prompt, the model and the wording.

Sources are the URLs an engine cited to build its answer. They are the clearest signal of what shapes your category, and usually most of them are not your own site. Changing what those pages say is often faster than changing your own.

Because the engines read what they trust, not what you would choose. Forums, review sites and aggregators appear constantly. An unfamiliar domain in your list is information rather than an error: it is where your category is being decided.

Engines that retrieve live pages at answer time will cite different sources as the web changes. Movement across a week is meaningful; a single day's difference usually is not.

Gap analysis lists the prompts where competitors are named and you are not, grouped by topic. It is the shortest path from a number to a task, because it tells you which specific question to go and win rather than that a score is low.

Earned covers what other sites say about you, which is where most AI visibility comes from. Owned covers signals on your own site. They are separated because the work is different: outreach on one side, technical and content fixes on the other.

ChatGPT, Claude, Gemini, Grok, Perplexity. All five are polled daily on the same prompt set so they can be compared directly.

Yes, under Settings, Engines. Turning one off stops new runs against it; the history already collected stays and remains visible in past periods.

Not every engine cites on every answer. Some answer from what they already hold rather than retrieving pages, and those answers carry no citation list. Perplexity cites most consistently, which is why its citation rate is tracked separately from its mention rate.

They read different sources and weigh them differently, so a brand can be prominent on one and absent on another. That disagreement is the reason to track them separately rather than as one blended score.

Only answers that actually mention your brand are scored, on a 0 to 100 scale, based on how the surrounding language describes you. Answers that never name you are excluded rather than counted as neutral, which would drag every score toward the middle.

Read it next to visibility, never alone. High sentiment on very low visibility means the few answers naming you are kind but rare. Falling sentiment at steady visibility usually means a specific source has started framing you differently, and Sources will show which.

Every competitor in your set is measured on the same prompts, on the same days, on the same engines. That is what makes share of voice meaningful: it is your share of a fixed pool rather than a number in isolation.

The 12-point audit checks the signals that decide whether an engine can parse and trust your site: structured data, FAQ schema, llms.txt and AEO signals, heading hierarchy, author bylines, content freshness, internal link health, image alt text, word count, Core Web Vitals, mobile friendliness, and HTTPS and redirects. Results are scored on three weighted axes: technical 0.30, content 0.35, AEO 0.35.

Under Settings, Billing. Your plan, renewal date and invoices are all there, and plan changes take effect from that screen.

Settings, Team. Invited people join the organisation and see the projects it owns, so there is no need to share one login.

Self-serve plans are listed on the pricing page, and there is a Managed plan where our strategists operate Noma on your behalf. If you are an agency running many clients, the agency page covers how that is priced differently.

From Settings, Billing. Access continues to the end of the period you have already paid for, and your data stays in place rather than being deleted at cancellation.

Settings, Notifications. Reports are sent on a schedule to the addresses you choose, which is usually how teams keep an eye on movement without opening the dashboard daily.

Yes. Each project is a separate brand with its own prompts, competitors and history, and nothing is shared between them. How many you can run depends on your plan.

Under the project's Settings. Changing the domain affects what the audit crawls, so the next audit will reflect the new site while past results stay attached to the period they were measured in.

A project holds a list of target markets, set during onboarding and editable in project settings. They shape the prompts Noma suggests and appear on the shareable report.

Every project has a public snapshot at a URL you can forward to someone without a Noma account. It shows your visibility, share of voice, sentiment, competitors and top sources. Anyone with the link can open it, so treat the link as the access control.

Yes. It is authenticated with an API key and returns the same data the dashboard reads, which is how teams pull visibility into their own reporting.

Settings, API keys. The key is shown once at creation and stored only as a hash, so it cannot be recovered later. Keys can be revoked at any time and stop working immediately.

Yes. Panels on the dashboard export to CSV, which covers most reporting needs and feeds any spreadsheet or BI tool. For anything continuous, the API is the better route.

Email hello@nurdd.club. If your question is about a specific number, include the project and the date range, since most questions of that kind are answered by looking at the exact prompts behind it.