An SEO report lands on your desk every so often. Dozens of items, plenty of red, words like critical and blocking. You paid for it, so the question is entirely fair: does this need to happen, and what does it cost us if we skip it?

The honest answer is that part of such a report is right, and more often than not a larger part no longer is. A lot of the advice comes from an era when search engines still counted words instead of understanding text. And many reports are produced by an automated tool: it lists everything that technically could be different, without weighing what it delivers. A list of 120 items does not mean you have 120 problems.

After reading this you can sort any piece of SEO advice into three categories yourself — already handled, worth picking up, or outdated — and explain why you deliberately spend no money on something.

This is not an argument against SEO. Good findability matters, good content matters more, and there is almost always ground to gain there. It is an argument against chasing numbers no applicant will ever notice.

Before you begin

Make sure you have three things at hand:

• The report itself, including who produced it and with which tool. Many reports come out of a handful of standard tools; that is not a disqualification, but it often explains why the items are not weighed against each other.
• Access to Google Search Console for your own domain. That is where the numbers Google itself measures about your site live — which is something other than what a testing tool predicts.
• Your site runs on the Booston website platform. The technology underneath is the same for everyone. What you have added on top — tracking, chat, video, separate landing pages — differs per customer, and that is where most of the differences in measurements come from.

What really drives your findability

For a site with jobs on it, it broadly comes down to four things: can a search engine find and read your pages, does it understand that there is a job on them, does the page load decently for a real visitor, and is there anything more valuable on it than on the rest.

The first three are technology, and they are built into the platform. The fourth is content, and that is and remains human work.

That is the foundation, and it is rarely the problem. What does differ per site is the set-up: how many pages you have, how you tie them together, which third-party scripts sit on them and — by far the most important — what the text says.

Advice that once worked and now does nothing

This is the hardest part of the conversation, because the advice was not made up. It did work, only in a different era. In the meantime search engines moved from "which page contains this word most often" to understanding whole sentences. That is what stripped the value out of exactly the tricks you still see in reports.

Three classics we keep seeing:

• Keyword density — a fixed percentage of the text in which your search term must appear. Google has repeatedly said there is no ideal density and that it is not a ranking factor. The only place where Google still mentions word frequency is its spam policies: as a penalty for keyword stuffing, not as a reward.
• Link sculpting — throttling internal links to steer link value. Google's then head of webspam explained publicly in June 2009 that this had already stopped working more than a year earlier: the throttled value does not go to the other links, it evaporates. That advice is seventeen years old.
• URL structures with categories and levels — putting jobs in a path with region and job family because it would score better. In its own documentation Google describes URL structure purely as a matter of discoverability and readability for people, not as a ranking factor, and has explicitly confirmed that flat and deep structures are not treated differently. Restructuring your URLs also almost always costs something: every old link has to be redirected and the history those pages built up partly starts over.

Why this pattern keeps repeating

Every new development produces a new list of tricks you "really have to do now". In 2009 that was the meta keywords tag, which Google was already openly ignoring at the time. Today it is llms.txt: a file that is supposed to register your site with AI search engines.

Google put an unambiguous sentence about that in its documentation on AI features: you do not need to create new machine-readable files, AI text files or markup to appear in the AI answers. Google publicly compared the file to that very same old meta keywords tag, for the same reason it does not work: it is a list the site owner fills in themselves, so it says nothing.

How to read a speed score

Google's speed test turns up in virtually every report, usually as a single number with a colour. That number is rarely explained, which is a shame, because it is quite easy to understand.

It is a lab measurement, not a measurement of your real visitors. The test loads your page once on a simulated, deliberately slow device over an artificially throttled connection. Two measurements in a row therefore often produce different outcomes — Google itself names routing, test device, browser extensions and antivirus software as causes. One screenshot of one measurement is not proof.

The number is also a weighted sum of five parts. In the current version, the time the page spends busy with JavaScript counts for 30%, loading the largest element for 25% and shifting layout for 25%. In other words: almost a third of your score is about how much functionality sits on your page.

What Google itself does with it

Google does not use your test score. What does count are the Core Web Vitals, and those are measured on real visitors in Chrome, over a longer period. You will find those numbers in Search Console, not in a testing tool.

And about their weight, Google is remarkably level-headed in its own explanation. There is no single composite number for page experience: "There is no single signal." And more importantly: Google always shows the most relevant result, even if the page experience of that page is sub-par. Speed is the tiebreaker between two equally good answers, not an entry ticket.

The 9.4 we can reach

A score of 9 or higher is achievable, and we have a page that genuinely reaches 9.4. Except: that page holds nothing but text and images. No video, no map, no chat, no tracking, no application form that thinks along with you.

So that score is reproducible, and the price comes with it. Virtually every feature you offer a candidate soon costs a couple of tenths. The question is not whether it could be higher — it always could — but whether you want to remove functionality that helps candidates in order to raise a score candidates never get to see.

We choose the best experience by default, because in practice that brings in more visitors and more applicants. If you would rather steer for the maximum score, that is entirely your right and we will help you — and we will make clear up front what you are giving up for it.

Where your budget does pay off

Once the technology is in place, most of the remaining gains sit in content, and in whether your job says something the other twenty do not. That counts double now that answers are increasingly summarised by AI.

What works there is, strikingly, exactly what works for ordinary search results: readable text in the page itself, a clear structure with headings, facts stated explicitly instead of implied, and crawlers that have access. For jobs that means very concretely: spell out the salary, the location, the hours, the contract type and the requirements. An AI answer only quotes what it can literally find.

Three places where that gain sits:

• Job texts that speak to your audience instead of repeating the job description — see Job texts with AI.
• Knowing who you are actually looking for and which words that person searches with — see Candidate audience analysis.
• Measuring what it delivers in applications instead of in positions — see Analytics.

This is also exactly the budget that often disappears into technical micro-optimisation. A monthly report on the length of your meta descriptions will not bring you a single candidate. Five well-written job texts will.

Where your SEO partner may well be right

There is something correct in virtually every report. Every platform carries technical debt: things that were once solved differently and could be done better with what we know now. That goes for us too.

The difference is in the weighing. We pick up what demonstrably does something for the visitor or for findability, and leave what only changes a checkmark in a tool. If an item turns out to hold water, it goes into the platform and everyone benefits — which is what you should expect from a platform, rather than a repair on your site alone.

So do send the report in. We are not out to prove your agency wrong. We are out to make sure your money goes to the things that work.

What we do with your report

Send the report to support@booston.io. We go through it and sort every item into one of three boxes, with the reason for each. No two reports are the same, and no two sites either, so we judge per item rather than per category.

Two things fall outside that:

• Content is custom work. Rewriting texts, restructuring pages, building a content plan: that is work, and we quote for it. It is usually also the part with the highest return.
• Do you want an item from "deliberately not" carried out anyway? That is often possible as paid custom work. The line we draw: if it only costs time and harms nothing, we build it. If it demonstrably makes your site slower or worse for candidates, we do not — not even for payment. You hire us for the result, not only for the execution.

Five questions to ask your SEO partner

These questions are not meant to corner anyone. A good partner answers them effortlessly, and then you know you are working with the right people.

1. Is this based on a lab measurement or on the data from my real visitors in Search Console?
2. What is the expected impact on applications — not on the score and not on the position?
3. Which year does this advice come from, and which current source from the search engine itself still confirms it?
4. How do we establish in three months whether it worked, and what is the agreement if it did not?
5. Which part of the measured load time comes from scripts added on the advice of a marketing or tracking party?

Frequently asked questions

Our agency says we need to score at least a 9. Is that true? No. There is no threshold below which your site stops competing. Google uses no test score at all, but the measurements on your real visitors, and weighs those only between results that are equally good on substance. Check in Search Console whether you are in the green; that is the measurement that counts.

Do we need an llms.txt file for ChatGPT and AI search engines? No. Google states explicitly that no new machine-readable files or markup are needed to appear in AI answers, and no major AI provider has committed to using the file. What does work is readable text and explicit facts on the page itself.

Our jobs are also on job boards. Is that duplicate content and are we penalised for it? No, there is no penalty for it. With identical text, search engines pick one version to show. If you want that to be your page, you help yourself most by making your own job page richer than the job board version: team members, employment terms, imagery, stories. See also Jobboard network and Job feed.

Can we change our job URLs to categories or regions? Technically yes, but do not expect better positions from it. Weigh it against the cost: every existing link has to be redirected and the history of those pages partly starts over. Do it if it helps your visitors, not because a report advises it.

The report says our images are not optimised. Have us check the item. Images are automatically converted to WebP and only loaded once they come into view. What you upload yourself does count: a 6 MB photo remains a large photo. See Images and files.

Need help?

Received an SEO report and want to know what in it actually matters? Mail support@booston.io or call 085 0044 215 and mention:

• the name of your environment;
• the report as an attachment, including who produced it;
• what you want to achieve — more applicants, more visitors or a specific position.

Would you rather go through the items together? Book a free 30-minute online demo — then we will walk through it together. Job Done!