SEO Intelligence

Every month, SEO Intelligence reads your Google Search Console data, works out what changed, and tells you in plain English what to fix first. No jargon, and every number can be traced back to the data it came from.

Overview

Search Console already knows how people find you on Google. What it does not do is tell you what to do about it. This service does that part:

• Pulls your search data every month and keeps a dated history of it.
• Compares this month against last month and finds what genuinely moved.
• Ranks the findings by how many clicks are realistically at stake.
• Explains each one in ordinary language and says what to change.
• Turns anything you agree with into a task for whoever does the work.

The four words to know

These are the terms Google uses, and the ones you will see on the Performance page and in every SEO article you read.

Connect your Google account

Go to SEO Intelligence → Settings, open the Google Search Console tab and choose Connect Google Search Console (once one account is connected, the button reads Connect another account). You need the permission to connect accounts, which admins have. You will be asked to allow read-only access to your Search Console data, and nothing else.

Before you are sent to Google, you are asked Who can use this Google Search Console account?: only you, the whole team, or specific people. Choose, then Continue to Google Search Console. Reconnecting an account later never changes that choice.

Leave every permission ticked on the Google screen. If the Search Console permission is unticked, the connection is refused and you will be asked to try again, because without it there is nothing to read.

You can connect several Google accounts. That is useful when different websites sit under different accounts. You need a real Google account that can open the website in Search Console.

Who can see what

Permissions decide whether you can do something at all. Sharing decides which records you can do it to. They work together, and you set the sharing yourself on each thing you add.

Three things carry their own sharing setting:

Each one is set to Only me, Whole team, or Specific people. With specific people you pick roles, and for each role either everyone in it or named members.

Everything that comes from a website follows the website: its performance figures, monthly reports, findings and recommendations. Everything that comes from a repository follows the repository, so its code audits are shared exactly as the repository is.

Sharing a website lets the people you shared it with see its search data even when the Google account behind it is private to you. They still cannot see or use that account anywhere else.

Add a website

On SEO Intelligence → Websites, choose Add a website. Pick the Google account, then pick the Website from the list, give it a Name, choose how much history to import and who can see it, and press Add website. The list comes straight from Search Console, so it only ever shows properties that account can actually open.

You do not type the address yourself. It comes from the property you chose, so the two can never disagree. A website you already track is marked Already tracked and cannot be picked twice. If no Google account is connected yet, the dialog says so and offers Connect a Google account.

How much history to import

When you add a website you choose how far back to import: 3, 6, 12, or all 16 months. Sixteen is as far back as Google keeps this data, and it is the default. The import runs in the background and takes a few minutes.

This matters more than it sounds. Several kinds of advice, such as spotting a page that has been quietly losing traffic, need several months of history before they can say anything at all. With only one month loaded, the first report will look thin. Pick a shorter range only if you want the first report sooner.

You are not stuck with the choice. Open the website and use Import history to pull more at any time. Months you already have are refreshed rather than duplicated, so importing again is always safe.

Tracking the searches you want to win

On the Performance page, under Searches you want to win, type the wording a customer would actually put into Google, such as best [what you do] near me, and press Track. Each search you add is tracked from then on, up to 100 per website. Adding and removing tracked searches needs the permission to add and edit websites.

For every tracked search you get where the site ranked in the most recent complete month, whether that is better or worse than the month before, and the best position it has ever reached. A move from 20 to 18 shows as 2 places better.

What this cannot tell you. Search Console only reports searches where your site actually appeared. If you rank far down for a search, Google reports nothing at all rather than a position, which is why a search can read Not seen even though the page exists. It also cannot tell you who is ranking above you.

Getting suggestions for a search

Beside every tracked search there is a How to improve action. It writes a short set of actions you could start this week: what to change on the page that already appears, or what to publish if none does.

The suggestions are written by AI from the figures Search Console reported for that search, and from the pages of yours that appeared for it. It is told never to invent a number and never to claim knowledge of who ranks above you, because it cannot see the results page. If something reads like a guess about a competitor, treat it as one.

Press Suggest improvements to write the first set. Producing suggestions uses AI credits, so the result is kept once written. Opening it again costs nothing. Use Suggest again when the numbers have moved and you want a fresh read.

How up to date the numbers are

Google publishes Search Console data about three days behind, so the most recent day you will ever see is roughly three days ago. That is Google's delay, not ours.

The Performance page keeps up with it. Each website is checked on its own schedule and the current month is refreshed up to the freshest day Google has, so the clicks, impressions, position and CTR figures, and the three charts, include the month in progress.

Open a website and use Check for new data to choose its schedule:

The card also shows when the website was last checked. Changing the schedule takes effect from the next check, and Import history refreshes the current month straight away whenever you want it sooner.

The charts run right up to today, so the most recent days appear here as soon as Google has them, exactly as they do in Search Console. Google revises the last day or two, so those figures can still move slightly after they first appear.

The monthly report is different on purpose. It only ever covers a month that has finished, so its findings compare like with like.

Pages, searches and countries for exact dates

On the Performance page, the Pages and queries and Countries tabs ask Google for exactly the days you pick, so a week, a fortnight or any custom range works, not only whole months. Choose 7 days, 28 days, 3 months, 6 months, 12 months, All, or your own dates, and turn on a comparison to set them against the period before or the same period last year. Each figure then shows its change.

• Pages and queries lists every page, or switch to Searches to list every search, with clicks, impressions, CTR and position. Find a page or a search by typing, and page through the full list.
• Blog posts narrows the list to your blog posts only. It appears when Blogs is active. Posts are matched by their public address, so set your Public blog URL in Blogs first.
• Pick a country to see pages or searches from that country only.
• Searches on a page row (or clicking the row) shows the searches that bring visitors to that page and where those searchers are.
• The Countries tab lists every country. Pages on a country row opens Pages and queries filtered to that country.

The same exact-date figures are on Blogs → Analytics, next to each post.

The growth board

This is the page you land on. It is one ranked list of what to do next, pulled together from everything the service knows about your website: the searches you nearly rank for, the topics you have no page for, the pages Google cannot see properly, and the pages your own site barely links to.

Each item carries a score from 0 to 100. Open any item and Why it is ranked here shows the actual numbers behind it: how many times those searches were shown, where you sit, how the clicks compare with what a position like that usually earns, and which way demand is going. Anything that had to be assumed is marked as assumed. Nothing is a black box.

An item stays on the board across months, so you can see that something has been worth doing since June and nobody has done it. Press Start it when work begins, then It worked or Did not work when it is done; Not worth it closes it for good.

Three tabs split the board: To do for the work, Already working for gains that need no work, and Closed for everything finished. Above the list, What changed shows alerts, such as a sudden drop in clicks or links that disappeared, and the most serious ones also arrive as notifications.

Content: what to write next

Your own search data tells us what people look for when they find you. We group those searches into topics, then check which topics your pages actually answer. A topic with real demand and no page is the clearest thing you can act on.

Pick one and press Write a brief. The brief is built from the real searches behind that topic, the questions people ask, the pages you already have that come close, and what your own page currently says. You get an outline you can hand to a writer. Writing a brief uses AI credits.

The Roadmap tab is simply the order you want to work through them in.

Technical: can Google see it

Four things, in the order that usually matters:

• Indexing - where your sitemap, your own website and Google disagree. A page in your sitemap that does not exist, a page on your site that is missing from the sitemap, a page published for months that Google has never shown.
• Sitemap - which sitemaps we found and how many pages each lists.
• Redirects - pages that take two or more hops to arrive, loops, and pages that point at a page that points somewhere else again.
• Speed - what real visitors to your website actually experienced, from the record Google keeps of it.

On page, and AI search readiness

Titles, descriptions, headings, image descriptions and the structured data Google reads. We only ask for the structured data that suits the page, so a blog post is never told to add product markup.

How you appear shows the result types you already earn in Google, measured, not predicted. If you never appear as a question-and-answer result on searches you already rank well for, that is worth a look.

AI search readiness scores how well a page is structured for an assistant to read and quote: does it answer in the first paragraph, are headings phrased as questions, is there structured data, is it dated, does it say who wrote it.

Internal links finds pages that earn clicks but that the rest of your own website barely links to, and names the specific pages that could link to them.

Competitors

Add the websites you compete with. We read their public pages, the way any visitor can, and match what they publish against the topics your own searches already show demand for. You get three kinds of gap: a topic they cover and you have no page for, a page of theirs that goes much deeper than yours, and structured data theirs carries that yours does not.

Did it work

The part most SEO tools skip.

Experiments. Name the pages you are about to change. We pick comparable pages you are leaving alone, matched on the kind of page and how much traffic it already gets, and then watch both. After at least three weeks we compare how your changed pages moved against how the comparable ones moved over the same days, and test that gap against a thousand reshuffles of which pages were which.

• It worked or It went the other way - the gap is bigger than reshuffling produces.
• No difference - it is not.
• Not enough to say - too few comparison pages or too few days. You get this rather than a weak number dressed up as a result.

Health is a month-by-month score across five areas. Every month shows how many of the five it could actually measure, because a score from five crawled pages is not the same as one from five thousand. Known Google updates in the period are listed underneath, which says when something changed at Google, never that it caused what your website did.

Conversions connects your other SoftSolz services. Switch on Forms, Newsletter, Blogs, Knowledge Hub chat, Leads or Deals, put a value on one result, and pages start being ranked by what they are worth rather than by clicks alone. A page that gets few visitors but turns them into customers then beats a page that gets many and turns them into nothing.

The monthly report

Open a website and choose Create report. It reads the month, compares it against the month before, and produces two lists:

• What to do, the actions worth taking, most valuable first. The first five are shown, because a month's work should be finishable; the rest are one click away.
• The data behind it, every observation the analysis made, with the figures that produced it. This is the evidence, not a second to-do list.

Anything that grew on its own appears under What went well instead, with no action attached. Each action also says what the clicks figure means, because the number can be clicks you could still win, clicks you gained, or clicks you lost.

A month can only be analysed once it has finished and had about three days to settle, so that its findings compare like with like. Create report always picks the latest finished month.

You do not have to remember: a report is also made by itself each month for every connected website, and the first one is made the morning after you add a website. Each month is analysed once, so pressing Create report for a month that already has a report says That month has already been analysed for this website.

What to do next

Each recommendation carries two things worth reading before anything else:

• Clicks at stake, roughly how many extra visits a month this is worth if you fix it.
• Effort, whether this is a quick win, some work, or a bigger job.

Together those tell you what to do on a Tuesday afternoon and what to plan properly. The list is already ordered so the best use of your time is at the top.

Use the tick to accept anything you agree with, or the cross to dismiss anything that does not apply to you. Dismissing keeps the record but drops it out of the way.

The kinds of advice

Turning advice into work

Any recommendation can become a real piece of work. Choose Create a task on the card, or tick several and press Create N tasks to create them together.

The request goes to AI Review first, where somebody approves it. Once approved, the tasks appear in Tasks and can be given to whoever does the work. Nothing is created behind your back.

A recommendation that already has a task waiting shows as Awaiting approval and cannot be sent twice, so clicking again by mistake costs nothing.

Sharing the report

Choose Download PDF to save the month as a report you can send to anyone: a colleague, a client, or an AI assistant you want to think about it with. It carries your workspace name and logo, the headline numbers, every action with the reasoning behind it, and the evidence table. The same download is available from the Download button on any finished report in the Monthly reports list, so you do not have to open a report to send it.

Choose Publish report to turn the month into a written summary instead. It goes through AI Review in the same way, and once approved it is saved to your Knowledge Hub, where anyone with access can read it and the chatbot can answer questions from it.

Links

The Links page covers the other half of SEO: who links to your website, and how the pages of your own website link to each other. It has four tabs.

Backlinks

Google does not let software read its links report directly, so for Google this starts from the export it lets you download. Bing can be connected directly.

In Search Console open Links, then under External links choose Top linking sites, click a site, and export its linking pages. That export has one row per link, which is what can be tracked here. It does not say which of your pages each link points to, so the import asks you; the home page is the default. The Top linking sites summary on its own is a list of domains and counts, so it is refused with a note explaining which export to take instead.

If a file's columns are not recognised, the import asks you which column is which: the page that links to you, the page on your website, the link text, or the date first seen.

In Bing Webmaster Tools open Backlinks and choose Export All. Or connect Bing under Settings on the Links page: paste a Bing Webmaster API key, pick the same website in Bing, and use Sync now to pull its links in without an export.

Each row shows what was measured about the link, not an estimate of what it is worth:

• Followed or nofollow, and whether the page marks the link as paid or as something a visitor posted.
• Where it sits on the page: in the writing, or in a menu, sidebar or footer.
• Whether the linking page is set to noindex, in which case the link is unlikely to count for anything.
• How many other sites that page links out to, which is how directory and spam pages give themselves away.
• Whether the page it points at on your website still answers. A link pointing at a page that returns 404 is worth fixing, and it is usually a redirect away.

There is deliberately no domain authority or spam score. Those numbers come from a paid third-party index; without one, the honest thing is to leave the column out rather than estimate it. What replaces it is the list above, which a score could never tell you because it comes from opening the page.

Every count says where it came from, for example 42 Search Console, 18 Bing, 31 we checked ourselves. A link nobody has opened yet reads Not checked, never live.

Check 25 links opens the 25 pages checked longest ago and records whether each link is still there. Once a day up to 200 more per website are checked on their own, and you are told when a link disappears or goes live.

You can also export everything as a spreadsheet, and once you have disavowed a link, Download disavow file builds the file for the links you want Google to ignore. A whole domain is only written as one line when every link known from it has been marked.

Prospects

A prospect is a page you would like a link from. Record the page, the page on your website it should point at, the wording you would like, and who to write to.

Add to an Outreach contact book puts every open prospect with an email address into a contact book called Link prospects followed by your website name, with the details filled in as merge fields, so you can write to them from Outreach with the consent and unsubscribe handling it already does. It needs Outreach, and permission to manage its contact books. Replies arrive in the Outreach inbox like any other conversation.

Remind me creates a task for you, due on the follow-up date, so it does not get forgotten (it needs Tasks). Check now opens their page to see whether the link has appeared; when it has, the prospect becomes a tracked link on its own. If a live link later disappears, the prospect moves to Link went.

Internal links

Read my website opens each page of your own website and maps how they link to each other. It reads up to 500 pages by default (you can change this in Settings, within your plan's limit), for up to 15 minutes. It obeys your robots.txt and waits between requests. If it stops early, it says whether it hit the page limit or ran out of time.

What it finds:

• Links that lead nowhere: pages of your own site you link to that no longer answer.
• Pages nothing links to: Google still shows them, but no page on your site points at them. One link from a relevant page is usually the cheapest fix in SEO.
• Pages buried deep, four or more clicks from your home page.
• Pages linked only with filler wording such as click here or read more.

The monthly report also gains a suggestion that only this crawl makes possible: when a search sits just off page one and another page of yours already uses that wording without linking across, it tells you which page to link from and what to say.

Mentions

Pages that name your brand without linking to you. Candidates are found by searching the web, then every one is opened and read before it appears: the phrase has to really be in the page. Anything that does not hold up is discarded rather than listed.

Turn one into a prospect to ask for the link.

Reading your website code

Search Console tells you how your pages performed. It cannot tell you why. Code answers that by reading the repository your website is built from and listing what the markup is missing: pages with no title or description, pages that block themselves from Google, missing structured data, and the files a site is expected to have such as robots.txt, sitemap.xml and llms.txt.

Connect GitHub once, either on the GitHub tab of Settings or with Connect GitHub on the Code page, then use Choose repository to pick the repository. You need a real GitHub account that can open the repository, and the permission to connect accounts. Only the files that render pages are read.

A website is often built by more than one repository, for example a marketing site and a docs site. You can connect as many as you need and each gets its own tab, audited and improved separately. You are asked which repository only when you connect it; after that everything you do on a tab applies to that repository.

Choose Run audit (later Re-run) to read the current state of a repository. The result is a score, a list of what is missing across the whole site, and a page-by-page breakdown ordered by how much traffic the fix is worth.

Letting us improve the wording

After an audit, choose Prepare changes and the missing wording is written for you: page titles, meta descriptions, canonical links, the mobile viewport tag, social sharing tags, structured data, and any of the four site files your repository does not have yet.

Pages that are failing quietly are rewritten too. A page can have a perfectly valid title and description and still be the most valuable page on your site to rewrite, because Google already ranks it and nobody clicks it. Those pages come from your latest monthly report, so the two halves of the service join up: the report finds them, the fix set rewrites them. Each one is labelled Rewrite the snippet so it earns the click and its reason names the position, how often it was shown, and the searches it was shown for, so you can tell it apart from a missing tag at a glance. Anything your team dismissed in the report is left out here as well.

For those pages the wording is written against the searches people actually used to reach them, not just the words already on the page. It still may only say what your page genuinely offers: a search never becomes a promise your site does not make.

Only wording and behind-the-scenes tags are ever changed. Nothing about how your site looks is touched. Stylesheets, layout, class names, components and framework settings are refused outright, and so is any change that would delete wording already on a page. If a suggested change would move anything visual, it is dropped rather than shown to you.

You see one row per file, with the reason for it. Untick anything you do not want, and choose Preview to see the exact lines that would be added.

Nothing is committed to your branch. Choosing Send N as a pull request, then Open pull request to confirm, puts the changes on a new branch and opens a pull request on GitHub. Your main branch is untouched until you merge it, and you can merge it whenever you like or close it and lose nothing.

Writing the wording uses AI credits, the same as every other AI feature. The tags and the site files are worked out from your own pages and cost nothing.

Connect an AI assistant or the API

If you already pay for Claude, ChatGPT, Gemini or another AI assistant, you can connect it to SEO Intelligence and let it do the writing. SoftSolz still does the measuring: it reads Search Console, works out what changed, checks your links, reads your code and opens the pull request. The assistant reads those numbers and writes the words.

A connected assistant can:

• Explain the findings in plain English, from the same figures you see on the finding card.
• Suggest improvements for a tracked search, using the position, impressions and clicks on file.
• Propose recommendations for a finished monthly report. They appear in the report like any other and wait for you to accept or dismiss them.
• Prepare wording fixes for your code and send them as a pull request, exactly as Letting us improve the wording does. The same rule applies: only wording and behind-the-scenes tags, never your design, and only the four site files can be created new.

Anything the assistant writes carries a Written by badge with the assistant's name, on the finding, the recommendation, the search suggestions and the prepared changes, so nobody mistakes it for something SoftSolz worked out.

How to connect

1. Open Developers in SoftSolz and copy the address from the Connected AI apps card.
2. Add it to your assistant as a connector, sign in, pick the workspace and tick the SEO permissions you want it to hold. Read websites, reports, findings, keywords, links and code audits lets it look; Start reports and write explanations, Write and decide recommendations, Connect repositories and run code audits and Propose and push code fixes let it write.
3. Ask in plain words: "Explain the findings in this month's report for acmebakery.co.uk", "What should I change to win the search sourdough starter?", "Prepare the missing meta descriptions and open a pull request."

The full walkthrough, including how to disconnect, is in Connect an AI assistant. A developer on your team can do the same from your own systems with an API key; the details are on the developer site.

An assistant sees what the member who connected it sees, so a website shared with Only me stays private to that member. An API key made by a developer sees only websites and repositories shared with the whole team.

Why the query numbers look small

You will usually see far fewer queries than pages. That is Google, not a limit here.

Google hides the search wording when too few people searched it, to protect their privacy. You still get the page and the visit; you simply do not get told what was typed. On a smaller website this hides most searches, because each individual one is rare.

Your page numbers are complete. If a website is ever large enough that some very low-traffic rows had to be left out, the report says so on the page rather than quietly rounding down.

How many websites you can track

Your plan sets how many websites you can track at once. Removing one frees the slot immediately. Connecting more Google accounts does not change the number.

See Billing and plans for what your plan includes.

Who can do what

In the role editor each permission is shown by its description, grouped under SEO Intelligence:

Managers get every permission except adding and editing websites, connecting accounts, code and deleting. Members can view only. You can change this for any role.

A permission applies only to the records shared with you. See Who can see what. On top of permissions, a website's competitors, experiments and results can only be changed by the member who added it or a super admin.

Set these per role under Members & access → Roles. See Account and access.

Test it step by step

SEO Intelligence reads real data from Google, so you need a real Google account that can open your website in Google Search Console. If your website is not in Search Console yet, add and verify it there first. Nothing in these steps changes your website.

1. Install the service. Open Marketplace, find SEO Intelligence and click Install. You need to be an owner or admin to install services.
2. Use an admin account. Admins hold every SEO permission. Managers cannot add websites or connect accounts by default (see Who can do what).
3. Choose where to test. The sandbox (the menu under your name at the bottom of the sidebar, then Switch to Sandbox mode) keeps its own websites and reports, but it still reads your real Search Console data, so many people test straight in their live workspace.
4. Connect Google. Go to SEO Intelligence → Settings, open the Google Search Console tab and choose Connect Google Search Console. Pick who can use the account, press Continue to Google Search Console, sign in and leave every box ticked. Back in SoftSolz a message says Account connected and the account shows as Connected.
5. Add your website. Go to Websites → Add a website, pick the Google account and the website, keep 16 months of history, and press Add website. A message says Website added and the website's page opens.
6. Check the numbers arrive. After a few minutes, open Performance. You should see clicks, impressions, position and CTR with three charts, matching what Search Console shows.
7. Create the first report. On the website's page press Create report. A message says Analysis started, and the report's status moves through Reading Search Console, Working out what changed and Writing the explanations to Ready. The explanations use AI credits.
8. Read it. Open the report. Its findings are under What to do and The data behind it, and the Growth page now has items on its To do tab.
9. Optional: read your website. On Links, open the Internal links tab and press Read my website. When it finishes, On page, AI search readiness and Internal links fill in.
10. Optional: code. To test the Code page you also need a real GitHub account with access to the repository your website is built from. Connect it, choose the repository and press Run audit. A pull request is only opened if you choose to send one.

For developers

Create a sandbox API key from SEO Intelligence → Developer with the Read websites, reports, findings, keywords, links and code audits permission, then list your websites:

curl https://app.softsolz.uk/api/v1/services/seo/sites \
  -H "Authorization: Bearer sk_test_your_key"

A 200 reply listing your websites means it worked. An API key only sees websites shared with the Whole team, so an empty list usually means the website is set to Only me. The SEO Intelligence developer guide covers reports, keywords, links and code fixes.

If something goes wrong