Seenvero
Features

Every section, and what it is for

Twelve screens, described as they behave today. Each one says what it is good for and where it stops, because the limit is usually the part that decides whether a number is worth acting on.

Last updated

Engines a project can ask, today

  • ChatGPT
  • ChatGPT (API)
  • Microsoft Copilot
  • Google AI Mode
  • Claude
  • Gemini
  • Perplexity
  • Grok
  • Meta Muse
  • DeepSeek
  • Google AI Overviews

Overview

Where you stand this window, and on how much data.

  • One sentence about you rather than a definition of a chart: how many of the window's successful answers named your brand.
  • Visibility, share of voice, citation rate, sentiment and average position, each with the change against the previous window of the same length.
  • The tracked brands beside you, and the ten most recent answers, newest first.
Why it helps

You can tell in a glance whether the week moved, and every figure states the sample it was computed over instead of floating free.

What it will not tell you

The denominator is successful answers only. An engine that returned nothing is a coverage gap, not a zero, and a day with no collection stays blank rather than being drawn as a drop.

Brands

In the app: Brands, and Brand Insights

Who you are measured against, and how engines talk about you when they bring you up.

  • Your own brand and the competitors you track, each with several websites and name aliases, so one company is one row rather than four.
  • Brand Insights splits every mention into recommended, listed, mentioned in passing or cautioned against, per engine, and shows the sentences behind the split — the flattering one and the damaging one.
  • Quick wins: the questions where you are close to being included.
Why it helps

You learn whether your problem is absence or treatment. Being in the answer and being the pick are different problems with different fixes.

What it will not tell you

Name and alias changes apply to answers collected afterwards; answers already stored are not reclassified, so a roster change does not rewrite last month.

Prompts

The questions you want measured, in your buyers' words.

  • Your prompt list, grouped into topics and tagged, with the country each one is asked from.
  • Per-prompt visibility for your own brand, so you can see which questions you lose rather than only an average.
  • Run a single prompt on demand when you do not want to wait for the next collection day.
Why it helps

The measurement is about the questions that matter to your business, not a generic keyword list somebody sold you.

What it will not tell you

These are the questions you chose. They are not observed customer searches and they are not search demand; no number on this site is a search volume.

Responses

In the app: Runs, and the page for a single run

The evidence itself: every answer, kept.

  • Every stored answer, filterable by engine, including engines the project no longer runs — an answer already given does not disappear.
  • A single answer in full, with how each brand appears in it, the citations it carried, and the searches the provider reported the engine ran.
  • Measurement details on every run: engine, model, collection method, country, timestamp, and whether a field was not exposed by the provider rather than empty.
Why it helps

Every number in the product opens the answer it came from, so a finding can be argued with rather than believed.

What it will not tell you

An answer is one sample at one moment. The same prompt asked again may answer differently, which is why the schedule exists and why a single answer is never presented as a trend.

Sources

In the app: Domains, URLs and Search Terms

The pages answers are built from, and whose they are.

  • Domains and URLs cited across the window, with retrieval counts, the type of page, and your own pages marked.
  • The search terms providers reported their engines ran on the way to an answer, when the provider exposes them.
  • Filters by window, engine and topic, with the same filter applied to the list and its denominator.
Why it helps

You see who is being read on the questions you lose — which is usually a more useful list than a competitor ranking.

What it will not tell you

A citation list is what the answer disclosed, not the engine's full retrieval, and reported searches come from the provider. Where an engine reports nothing, the product says so instead of guessing.

Gap Analysis

The answers that name competitors and not you, and the pages they used instead.

  • Cited-instead: the domains and pages that appeared in answers where your brand did not, by domain, URL or host.
  • Your own site read in an answer that never named you, shown apart — the page was good enough to consult and not good enough to be credited.
  • An on-demand check of the top cited pages from the last 30 days: it downloads the page and reports whether your brand's text is on it, who owns it, and when it was checked.
  • A confirmed opportunity can be saved straight to Actions with its evidence attached.
Why it helps

Instead of a score telling you that you are behind, you get the specific pages standing between your buyer's question and your name.

What it will not tell you

Checks read bounded downloaded HTML and do not render JavaScript. “Not found” means not found in the inspected text, and a page as it stands today is not proof of what an engine read last week. Checks are rate limited per project and normally refresh after a day.

Page Audit

Whether a page can be fetched at all, and what it tells AI crawlers.

  • Submit a URL: status and response time, the final URL after redirects, canonical, meta robots, X-Robots-Tag, title, meta description, whether structured data is present, and the size of the HTML inspected.
  • robots.txt permission for each AI crawler — OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Bytespider and bingbot, the token that decides what Microsoft Copilot can see — evaluated against the final URL, path and query.
  • Optionally compare robots.txt across the project's own and competitor domains in one table.
Why it helps

It separates a content problem from a plumbing problem before you spend a month writing. A page a crawler is refused is not a content problem.

What it will not tell you

HTTP only: no browser rendering, no model calls, and an unreachable page or a robots file returned as HTML stays “unknown” rather than being read as permission. Allowed means robots.txt permits it, not that any crawler came or that any engine used the page.

Optimizer

In the app: Content Optimizer

Turn a gap into specific words on a specific page.

  • Improve a page: the optimizer inspects the page and returns small, ready-to-copy section edits — the section, the proposed text, the reason, a priority and one provider-reported target query each.
  • Create a topic brief: an outline built from provider-reported queries and the approved facts on your Profile, with unsupported brand claims labelled for confirmation.
  • Every quoted original must actually exist in the inspected page before a result is stored, and a partial extraction is shown as partial rather than dressed up as a full-page comparison.
  • Copy an edit, download the result as Markdown, or save a section to Actions.
Why it helps

The output is an edit somebody can paste, grounded in evidence you can open, rather than advice to “improve your content”.

What it will not tell you

Suggestions need a human review before they go live. The page score is a heuristic for prioritising work; it is not a probability of being cited, and nobody can offer you one.

Actions

The work, with an owner — and the retest afterwards.

  • Tasks with an owner, status, priority, effort and the URL they concern, created from a recommendation or by hand.
  • Typed evidence chips linking back to the prompt, answer, cited URL or reported query that argued for the work.
  • A follow-up record: the date a change went live, what changed, the affected URL, the tracked prompts, and a 7, 14 or 28-day comparison window.
  • Regenerating the analysis never erases a decision or duplicates a task.
Why it helps

A finding survives the meeting it was found in, and the next person to open it can see why it exists.

What it will not tell you

The comparison only runs on identical prompt wording, engine, geography, model and collection method on both sides of the date; anything else is reported as insufficient comparable data, with the coverage counts. What it shows is an observation about two samples, never proof that your change caused it.

Report

The window written down, for somebody who was not in the tool.

  • The same window as the Overview, with its sample size, coverage by engine, the collection methods and models behind the answers, the brands table, planned work and what was completed.
  • CSV downloads carrying the same dates, labels and outcomes as the screen.
  • An admin can create a limited client summary on a link that expires after 7, 30 or 90 days, and revoke it. Owner names, private notes, prompts and full answers are not in it.
Why it helps

You can hand somebody a number and the sample it rests on in the same document, which is the only way a number survives a sceptical reader.

What it will not tell you

A report describes the answers that were collected in its window. It is not a statement about every buyer, and it says how many attempts failed rather than quietly dropping them.

Profile

What is true about your brand, signed off once.

  • Your brand's context, industry and attributes, used when the product needs to know what you actually do.
  • Approved facts: up to thirty statements, each with a source URL, plus your writing preferences, recorded with the time they were reviewed.
  • The optimizer may only use facts from this list, and checks each one back against it before showing the fact, its source or its review date as evidence.
Why it helps

Generated copy repeats claims you have approved and cites where they came from, instead of inventing a certification you do not hold.

What it will not tell you

Facts are yours to keep current. Nothing here is verified by us, and a stale fact with a source URL is still a stale fact.

Collection

What was asked, what is queued, and what went wrong.

  • Scheduled collection every day, with the day's attempt allowance and the work still waiting.
  • The latest attempts with their outcome: queued, running, collected, or one word for a provider refusal, an expired lease or a cancelled attempt.
  • Pause collection, or request a manual run; times and daily limits are UTC.
Why it helps

You can see whether a quiet week was a quiet week or a broken pipeline, which is the difference between a real finding and a false one.

What it will not tell you

An attempt that produced no answer is not retried automatically and is never counted as a measurement. Three collection days a week is a sample of the week, not a continuous recording.

Also in the application

Recommendations groups the window’s findings into priorities, quick wins and watch-outs, and is where most Actions start. Crawl Insights reads AI crawler traffic from a connected CDN account. Settings holds the project’s engines, schedule and allowance.

See it working before you decide

The sample report walks one invented brand through a full window. The free check is real and runs on your own page.

Check my page for freeSee a sample reportRead the methodology

Free page check

A free technical check of one page: whether it can be fetched, what it tells AI crawlers, and what its markup is missing. No account, no card. It finds obstacles on your page rather than measuring what assistants say about you.

Check my page for free

Sample report

A full fortnight of measurement on an invented brand: the answers, the brands they name, the pages they cite, and the work that followed. It shows you the evidence before you spend anything.

See a sample report

7-day trial

Your questions, asked on a schedule and monitored over time: every answer stored with its sources, the brands it names and how they move. The Starter trial runs 15 prompts on 1 project across 3 engines. A card is required so the subscription can begin on day 8; cancel before then and nothing is charged.

Start 7-day trial