Most agency case studies tell you a client's traffic “improved.” We're going to show you the actual Search Console numbers instead.
The short answer
One page, 61 to 1,186 impressions
We took one client page from 61 to 1,186 impressions in a single period — an 1,844% increase. Average position improved from 18.8 to 11.2, and the queries we had prioritised — the ones that match the client's actual buyers — moved onto page 1. We also found evidence the content is already being cited inside AI-generated search answers.
| Metric | Before | After | Change |
|---|---|---|---|
| Impressions | 61 | 1,186 | +1,844% |
| Average position (all queries) | 18.8 | 11.2 | −7.6 positions |
| Priority queries | Page 2 | Page 1 | Target audience reached |
| AI-prompt queries in top 3 positions | 0 confirmed | 3+ confirmed | New signal |
The client
An unglamorous starting point
HDS Learning is an experiential leadership and corporate simulation training company based in Dubai, serving corporate clients across the UAE and beyond. When we started working on their SEO, the starting point was not glamorous: no Google Analytics or Google Tag Manager was actually live on the site, despite both having been created in the account. The property existed. The container existed. Neither had ever been published to the live pages. That gap alone meant the client had been making decisions without real behavioral data for an unknown period of time.
We don't say this to embarrass anyone — we say it because it's a more common starting point than most agencies admit, and because the fix mattered more than the failure.
It's worth being specific about why this gap exists so often: a GA4 property and a GTM container can be created inside an account months before anyone connects them to the actual website. Someone sets it up, intends to install it later, and “later” never comes. From the outside, everything looks configured — the account exists, the IDs are real — but the site itself is sending zero data. The only way to know for sure is to check the live page directly, not the dashboard.
The method
How we found it — the method, not just the result
Before recommending a single change, we pulled the raw Google Search Console export directly — not a summarized dashboard view — and cross-referenced every page against the site's existing content inventory. This is a deliberate sequence, not a one-off check: identify striking-distance queries first (positions 11–20 with real search volume, since these convert to page-1 traffic with the least additional effort), check for keyword cannibalization between pages competing for the same intent, and only then prioritize by expected impact versus effort. Two findings stood out immediately from that process.
A page already breaking through, if we pushed it
The page /business-simulation-training/ was sitting at an average position of 18.8 with 61 impressions in the prior 28-day period. In the following period, impressions jumped to 1,186 — a 1,844% increase — while average position improved from 18.8 to 11.2. That average still sits on page 2, and it is worth being precise about why: it covers every query the page ranks for, including a long tail we were not targeting. The queries we had prioritised are the ones that reached page 1 — and we prioritised them because they match the client's target audience, not because they had the biggest volume. A page-1 position on a query an actual buyer types is worth more than a page-1 position on a query that merely gets traffic. This wasn't a page we built from scratch. It was existing content that had been under-optimized relative to real demand already visible in the data.
Proof the content is already inside AI-generated answers
When we reviewed the client's Top Queries report, we noticed something unusual: several of the queries driving impressions weren't keywords in the traditional sense — they were full AI prompts.
Query, as it appears in Search Console
in 3–5 sentences, describe how training simulations should be calibrated for [company name]: capture the unique tensions, strategic pressures, and cultural dynamics that should shape realistic decision-making scenarios for its leaders
Queries like this one were appearing in Search Console, with the client's page ranking in position 2 and pulling hundreds of impressions on each variant.
This matters because of how Google's own documentation describes AI features in Search: according to Google Search Central's guidance on AI features, pages that appear inside AI Overviews are counted within a site's normal Search Console performance data, under the standard “Web” search type — and Google has stated that clicks originating from AI Overview citations tend to correlate with higher-quality engagement, since users are more likely to spend additional time on the site after clicking through. In other words, this wasn't a vanity metric. It was a real, measurable signal that the client's content was being treated as citable source material by Google's generative search layer — not just indexed, but referenced.
The work
What we did about it
We didn't treat this as a one-off win to screenshot and move on. Once we identified the pattern, we applied the same cross-referencing method — Search Console data against the existing page inventory, checked for keyword cannibalization, checked for striking-distance opportunities (queries sitting in positions 11–20 with real volume) — to find the next candidates for the same treatment. That process surfaced additional opportunities sitting just outside page 1, including a related query cluster around executive simulation training still sitting around position 20, with volume that justified prioritizing it in the next sprint.
We also fixed the underlying measurement gap. Once we confirmed the client's GA4 property and GTM container existed but had never been connected to the live site, we installed the container, verified real-time data was flowing, and built a conversion event to track clicks on contact-intent actions — so that from this point forward, decisions about what to double down on are based on actual user behavior, not just impressions and position data.
We also took the chance to correct a detail that trips up a lot of GA4 installations. Google retired the standalone “GA4 Configuration” tag in September 2023, replacing it with the unified “Google Tag” model — but a large share of the setup tutorials still online were written before that change and walk you through a tag type that no longer exists. We caught this because we validated every step against the live Tag Manager interface rather than a memorised checklist. It is a small detail, but it is the kind of detail that determines whether a client's analytics are actually trustworthy six months from now, not just on the day they are installed.
For buyers
What this means if you're evaluating an SEO agency
Most of what we've described here isn't exotic. It's the product of treating raw data as more reliable than a dashboard summary, checking assumptions against the live site instead of the account settings, and being willing to say “this isn't working yet” instead of reporting a task as finished because the underlying pieces technically existed. Before hiring an agency, it's worth asking directly: can they show you the specific query, the specific page, and the specific before-and-after number behind any claimed result? If the answer is a general percentage with no underlying data to inspect, that's worth treating as a yellow flag, not a green one.
The standard
Why this is the differentiator
Anyone can tell a prospective client “we'll improve your SEO.” Very few agencies can show you the specific query, the specific position, and the specific percentage increase — let alone show you evidence that a client's own words are already being surfaced inside an AI-generated answer. That's the standard we hold our own work to, and it's the standard we think this industry is moving toward: not promises about rankings, but verifiable, client-specific proof.
This also isn't a result we're treating as finished. The same striking-distance analysis that surfaced this page also flagged a related cluster of queries still sitting outside page 1, which means the next reporting period has a defined, prioritized target rather than a vague hope that “more content” will help. That's the difference between SEO as an ongoing measurement discipline and SEO as a series of disconnected projects — and it's the reason we report on real Search Console numbers every cycle, not just the wins.
FAQ
Frequently asked questions
What does it mean if my content shows up in Google's AI Overviews?
It means Google's generative search layer has determined your page is a reliable enough source to reference when answering a related query. According to Google's own documentation, this traffic is counted in your normal Search Console data — it isn't a separate, invisible channel.
How do I know if my site is already being cited by AI systems?
Check your Search Console Top Queries report for entries that read like full sentences or instructions rather than short keywords. If you see prompt-like phrases with strong average position and no clicks, that's often a sign your content is the source, not just a search result.
Does a jump like 1,844% impressions mean 1,844% more traffic?
No — impressions and clicks are different metrics. In this case, the position improvement (18.8 to 11.2) is what makes the impression growth meaningful: more people are now seeing the page in search results at all, which is the precondition for clicks to grow next.
Why wasn't analytics installed if the account already existed?
This is more common than most agencies will admit. A GA4 property and GTM container being created doesn't mean they were ever published to the live site. It's worth checking this directly — via a tool like Tag Assistant — rather than assuming it's been done.
How long did it take to see these results?
The impression and position changes shown here reflect one full 28-day reporting period after the page was re-optimized, compared against the prior 28-day period. We don't promise a fixed timeline for every case, because it depends on how much existing demand a page already has in Search Console data before any changes are made — this page had real underlying search volume waiting to be captured, which is exactly what striking-distance analysis is designed to find.
Ready to see what's actually happening in your own Search Console data?
We'll pull the raw export, find the queries sitting just outside page 1, and show you the specific pages worth fixing first.
Book a strategy call
