Scan, gaps, rewrite, re-check

You type in a domain. No keyword list to upload, no competitor list to assemble, no tags to set up. The first run does three things at once.

• Finds who you actually compete with, taken from the sites ranking for the same terms you already rank for, not from a list you guessed at.
• Pulls every keyword the domain ranks for, with its current position, monthly search volume, how hard the term is, and what someone searching it is trying to do.
• Crawls up to 100 of your pages and keeps the parts a rewrite needs: titles, meta descriptions, headings and body copy.

One more check runs alongside those: whether an assistant can read your site at all. Is the content there in the HTML or does it appear only after JavaScript runs. Do your pages return sane HTTP responses. Is the document structured well enough for a machine to follow. That answer matters before anything else, because a page that loses because a crawler never saw it is a different problem from a page that loses because the copy is weak, and the fixes have nothing in common.

Measurement needs something to measure. Beseen turns your domain and its keywords into 25 prompts a buyer would plausibly type into an assistant. Not keyword strings. Full sentences with a purchase sitting behind them: the “best tool for X” question, the head-to-head comparison, the “what should I use instead of Y” question people ask when they are already unhappy with something.

Beseen generates these instead of asking you to write them, for a dull reason: almost nobody sits down and writes 25 good ones, and the sets people do write tend to flatter the brand rather than describe the market. Everything after this step rests on the prompt set.

Each prompt goes to ChatGPT, Claude and Perplexity. Beseen records the answer, which brands got named in it, and which pages got cited as sources. Perplexity lists its sources on screen, so for those answers you can open the exact page that beat you and read it.

The same run re-checks your Google positions, so both scoreboards move on the same day and you are never comparing this week’s citations against last month’s rankings.

How much gets checked, and how often

• Free: one property, 10 prompts against one assistant, weekly, plus one trimmed-down run of the site each month.
• $39 a month: three properties, 25 prompts across all three assistants, weekly.
• $79 a month: ten properties, 100 prompts across all three assistants, weekly.

Weekly is deliberate. Citations do not move day to day, so daily checking mostly buys you noise at seven times the running cost.

What comes back is not a visibility percentage. It is a list of prompts where a competitor was named and you were not, each one showing who got named instead of you and which of their pages the assistant used to answer.

Then Beseen works out why those pages won. Usually it is something structural rather than clever: they answer the question in the first paragraph instead of the fourth, they have the comparison table you do not, they put the brand name next to the claim so a citation has something to attach to.

A smaller version of all this runs without a signup. See exactly what the free scan checks.

Beseen only rewrites pages where a rewrite can realistically change the outcome. Where your page currently sits decides what happens to it.

• Positions 1 to 3: left alone. There is more to lose than to gain, so nothing is suggested.
• Positions 4 to 15: rewritten first. A better title and a rewritten opening can genuinely move these.
• Positions 16 to 50: the copy is not the bottleneck. You get missing sections and a structure to add, not new wording.
• Past 50, or not ranking at all: no amount of rewriting saves it, so it goes to the list of pages worth writing from scratch.

The parts it changes, in order of leverage

• The title tag. The most ranking movement per character changed, and the line an assistant is most likely to read as what the page is.
• The meta description. Judged on whether a human clicks, not on whether a keyword appears in it.
• The H1, which on most sites still describes something other than the term the page actually ranks for.
• The first 100 words, where both Google and the assistants decide what this page is about.
• The H2 structure and the sections that are missing, which is the content gap written out as an outline.
• Internal link anchor text, a free relevance signal that most sites spend on the words “click here”.

Every suggestion carries its reason

You always see the current version next to the proposed one, with a sentence you can check yourself: “this page sits at position 7 for a term with 2,400 searches a month and its title does not contain that term.” If the reason does not hold up, ignore the suggestion.

The suggestions that come from the citation half read differently from ordinary SEO advice, because they are about being quotable rather than being ranked:

• “Your intro buries the answer in paragraph four. Lead with it. That is the prompt you are not cited for.”
• “Every page cited for this prompt has a comparison table. This page has none.”
• “You answer the question but never name the brand near the answer, so the citation credits the topic instead of you.”

What it will not do

• It preserves your voice. It does not hand back copy that reads like a different company wrote it.
• It never gives you a keyword density target, because that is not a thing.
• It caps at 20 pages per run. A rewrite you have not read is not a fix.
• There is no apply-all button, on purpose.

Some prompts you lose because you have nothing that could answer them. Comparison pages, “best X for Y” lists and alternatives pages are what assistants reach for most often, and those are usually the ones missing.

For each of those you get the prompt it targets, the angle, the structure the page needs, and a prompt you can paste into whatever tool you already write in. The hard part is knowing which page to write, and that comes out of the list of prompts you lose, not out of the prose.

Output arrives in two forms. The finished copy, ready to paste. And a prompt written for a coding agent, so if your site lives in a repository you can hand the whole change over and review it like any other pull request. The structural findings work the same way: a page that fails a readability check for machines comes back as a prompt describing what to change and why.

This is a deliberate limit, not a missing feature. Writing to somebody’s production pages needs write permissions, a rollback story, and a support process for the day an automated change goes wrong, and none of that makes the answer more correct. The re-check does not care who applied the fix.

A week later the same prompts run against the same assistants, and your positions get pulled again. Every run is compared against your first one, so you can see which page changed which prompt, and which change did nothing at all. That last part is the one other tools leave out.

One honest caveat: citations move on the order of weeks, not days. The check right after you ship a rewrite will often look identical to the one before it. What you are watching for is a prompt that named three competitors and now names four, with you in the list.