SEO PulseBeta

Paste the client's website

One field, one click. Nothing to connect, no plugin on their site, no access to ask for — which is why you can point it at a prospect as easily as a client.

Every page comes back checked

The whole site is crawled from its own sitemaps and navigation, and each page gets the same checks a search crawler applies. You get a sortable table of the site with the pages worth a closer look already ticked, each with the reason it was picked.

Read the pages that carry the weight

Tick the pages worth it and each one is read in full against its own measured data — one named cause, the evidence behind it, the fixes ranked by effort, and the corrected title, description and heading ready to paste.

Hand the client the report

One click turns the whole scan into a white label SEO report in your agency's name — every page diagnosed, the full crawl ledger, and a recommended-actions sheet designed to detach. Download it or save it as a PDF.

Let Claude make the fixes

Connect the client's platform once — WordPress, Webflow, Wix, Framer, or the site's own code — and Claude applies the changes: each one shown first, written as a draft, and written down. You publish.

Scan again, and bill for it

The next scan checks every change against the live site. The Work Done report lists what was fixed, when, and whether it held — the page you send with the invoice.

Every page discovered from the site's own sitemaps and navigation gets crawled and checked — a hundred pages or a thousand. The table shows all of them, sortable by what each page earns, with near-identical template pages grouped so you tick one instead of forty.

No connection, no verification, no access request — paste a URL and go. That's what makes it a prospecting tool as much as a client tool: you can audit a business that has never heard of you.

Online store, local business, B2B or content site — detected from the site's own signals and stated up front so you can correct it, never asked for in a form. The audit is graded the way that kind of site is actually judged.

Every check below is a fact. The verdict above them is a judgment — what this site's structure is doing to this business, what it is failing to do, and what that is costing them — written after reading everything the crawl found, not stitched together from the ticks. A list of ✗ tells a client what is broken. This tells them why it matters, and it is the paragraph they read first.

Four questions asked of a local site the way a first-time visitor asks them: who is this, where do they work, what do they actually do, and how do I reach them? Each one answered with the evidence — how often the town is named, whether the phone is where a thumb can find it, and whether the number on the site is even the same one on their Google listing. That last mismatch is quiet, common, and it costs them local rankings.

Performance, accessibility, best practices and SEO, scored by Google's own Lighthouse — and where Google holds enough field data, what real people on real phones actually experienced loading the site. A client will argue with your opinion of their site speed. They don't argue with Google's.

Server- or client-rendered, and how much text a crawler sees before scripts run. A site that looks fine to you and nearly empty to Google is the finding that explains everything else on the page.

llms.txt, and where each AI crawler stands — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot — allowed, blocked, or simply never addressed. Most sites have never made that decision on purpose.

Sitemap, robots.txt, canonical tags, schema types, heading order, alt-text coverage, link previews, marketing pixels, review display. Each one a card in plain words — a grade tells a client nothing, a list of ✗ tells them what to do.

Not a checklist dump: each read page gets one primary cause with a confidence score, a diagnosis in two or three sentences, and the measured evidence behind it — its own checks, its keyword rows, and its Search Console lines where they exist.

A corrected title tag and meta description written to Google's display limits in the brand's own voice, an H1, and the schema type worth adding — each with a copy button. The fixes above them are ranked and tagged quick, moderate or involved.

Optional per read: real keyword rows with real monthly volume and difficulty, chosen from the index and from the queries Google already shows the page for — net-new targets, climbs from positions 4–30, and queries earning impressions but no clicks.

Every keyword the site ranks for is tagged by what the searcher wants — ready to buy, comparing options, looking for a brand, or just reading. A site can rank for thousands of terms and sell to almost none of them; this is the screen that shows it, and the list of buying-intent terms they could win now is the scope of the work.

Connect the client's Google and every figure upgrades to their own measured clicks, impressions, CTR and positions — per page, not just sitewide. One brand kit connects to one client's Google account, so nothing crosses between clients.

A local client's keywords get re-priced where they can actually be hired, not nationwide. Measured 44–67× apart on real local sites — which means a national figure can turn a keyword nobody local searches into the centrepiece of a strategy. Both numbers are shown, always, and the one you quote is the reachable one. It also returns the month demand peaks and the date content has to ship by to rank for it.

The competitors this site already ranks for, named. A rival's brand sitting inside your client's keyword set is not a curiosity — it is the sentence that opens the call.

Every audit now opens with the link profile: how many sites point here, how many of those actually pass authority, and a spam score for the whole thing. Free, on every scan, before anyone clicks anything. The story is usually in the mix rather than the total — a business can have fifteen hundred sites linking to it and four hundred of them are free blog pages.

Link checkers report a page as dead when a site simply refuses robots — one client came back with 1,906 broken links including their own homepage, which was working perfectly. Because we crawl the client's entire site in the same run, we can check each one against pages we reached ourselves. Nobody else holds both halves, so nobody else can tell the difference.

Every tool shows you a link. None of them tell you a national paper wrote that piece in 2012 and has never mentioned the client since. We read the publication date out of the page's own address, so an eleven-year-old write-up that still sends traffic shows up as what it is: the warmest outreach call an agency will make all week.

Worth a phone call, and worth removing. Publishers who wrote about them years ago and still link sit on one side; spam networks, throwaway hosting and sites no client wants their name beside sit on the other. A source never appears on both, because an agency cannot court a publisher and disavow it in the same afternoon.

Optional, priced per page, and the part nothing else does. You get the sentence that names the client — verbatim, not paraphrased — plus who wrote it, when, whether the article is even on the subject, and anything a client would not want their name beside. One diner turned out to be listed as a venue on an adult dating site. A DR81 link, on every dashboard in the world.

Roughly a quarter of publishers refuse automated visits, and it is disproportionately the big ones — the national papers you most want to read. Those are never billed. You get them back as a list of links so you can open them in a browser yourself, and the report says how many were skipped rather than quietly implying it read everything.

A ten-branch business audited as one site gets an average that describes nobody. Every location the site has a page for is found, with the keywords each one already carries and its best position — so you can see at a glance which branch is carrying the business and which one nobody can find.

The finding that sells the job: a location the business talks about but never gave a page cannot rank locally at all. It is sitting there in plain sight on every chain website, and no other tool hands an agency that sentence.

Paste the services straight off the client's Google Business Profile and each one is judged against what the site's pages are genuinely about. Three answers come back, and the middle one is where the money is: covered, no page at all, or — the one nobody catches — a page that talks all around the service and never once names it, so Google has nothing to match. That is an hour's edit to a page they already own, and no keyword tool will ever hand it to you.

For a branch worth digging into: 49 real Google Maps searches on a grid around its actual address. Google shows different results depending on where the searcher stands, so "they own the map pack downtown and vanish three miles north" is a picture a client feels — and one number never was.

Those 49 searches return every business ranking at every point, each with its own complete map. Put the client head to head with any of them, swap them in and out as you work through the account, and download each comparison as its own branded image to drop straight into the deck.

A restaurant and a plumber are not the same map. Walk-in keeps the grid tight around the address (about nine miles across) because people come to them; service area spreads the same 49 points across thirty miles because they drive to the customer. It changes where the searches happen, not how many. The right one is picked for you from the trade, and you can change it before anything runs.

Five questions written for that specific location — city and state in every one, because an assistant has no idea where you are — asked live to ChatGPT, Claude, Gemini and Perplexity. You get which of them named it, on which question, and the answer QUOTED. Expect most local businesses to score low today; "you're named by two of the four, let us get you into all of them" is a project with a finish line. Add it to any branch you dig into.

The check here answers one question about one branch: when someone asks an assistant for what this business sells in this town, does this location get named? When a client wants their whole brand measured that way, AI Visibility Pulse goes further — fifty questions, share of voice against competitors you name, which sources each assistant is reading, and how all of it moves month to month. Use this one inside a local audit; use that one when AI search is the conversation.

The client's logo in the masthead, your agency's mark top-right and your name in the footer. Executive verdict, scoreboard, site health read and graded in words, the services they advertise on Google set against the pages they actually have, where their demand really is, every single term they rank for and every term worth going after next, which of those keywords actually buy, every diagnosed page in full dress, the complete crawl ledger, and the method — download the HTML or save it as a PDF.

Every location you measured gets its own page in the report: its map beside the rival winning that map, the searches bringing people to it priced at ITS market, and the assistants' answers printed in full. On a ten-location client this runs to dozens of pages of measured work — the kind of audit that takes a person days, and reads like it.

Scan again after the fixes ship: the same pages, the same checks, and every number now has a before. That movement is the story a retainer is built on, and the report's last sheet is designed to detach as the client's action list.

Connect the client's platform once and Claude applies the scan's fixes: the new title, the shorter description, the keywords in the right places, the headings in order, the schema. Each change is shown before it's made and written as a draft. You publish.

WordPress and Webflow connect directly. Wix and Framer too. If the site is code — Next.js, Astro, plain HTML — Claude edits the repo. Shopify products go through your store connection. Squarespace can't be written to yet, so you get the exact text to paste.

Which platform, which connection, where the code lives, how a change goes live. Recorded once, read every time — so a teammate opening the client next week doesn't start from zero, and never gets asked for a password.

Open any page in the scan and see what's been fixed on it, when, and what the latest scan said: confirmed with the date, waiting on the next scan, or still failing. The page table shows the running tally.

Beside the client report sits a second button. It builds itself from the record of fixes: what was wrong, what changed, the date, and whether the next scan confirmed it — grouped by kind, then page by page, with every change ever made to the site at the back. Pick the period to match the invoice.

A fix is marked confirmed only when the next full scan finds the problem gone from the live site. Still failing is said out loud. That's what makes confirmed worth something to a client.