Reading Search Console without the wrong conclusion

Search Console occupies a unique position among the things discussed under this subject: it is free, it comes from Google rather than from a vendor's model, and it reports what actually happened to your own site. Every other source of search performance data is either a paid estimate or an inference from your own markup. This one is measurement, and it costs nothing beyond verifying that you own the property and connecting access.

That combination makes it the most valuable free resource in the field and also the most misread, because accurate numbers presented plainly invite confident conclusions, and several of these numbers do not support the conclusions people draw from them. The report is not lying and nothing is broken. The metrics are simply defined more narrowly than their names suggest, and the gap between the name and the definition is where decisions go wrong. What follows is the four places that gap does the most damage.

This is the metric that causes the most trouble. Average position is not where you rank. It is the mean of the positions your result occupied across every impression in the period, which makes it a summary of a distribution rather than a location. A page appearing at position 3 for half its impressions and position 40 for the other half reports an average around 21 — a figure describing a state the page was never actually in.

Several consequences follow. The metric moves for reasons unconnected to your page: appearing for a new query where you sit low drags the average down while adding traffic, which reads as decline on a dashboard and is growth in reality. It moves when a competitor's page is removed. It moves when the mix of queries shifts seasonally. And it cannot be compared meaningfully across pages that appear for different numbers of queries. The defensible uses are narrow: watch it for one specific query over time, where the average is over a coherent set, and treat a site-wide average position as roughly meaningless. If you want to know where you rank for a query, the honest answer is that this number does not tell you, and no free source does.

An impression is recorded when your result appears in a results page the user saw. That is a lower bar than the word suggests. It does not mean the searcher read your listing, noticed it, or scrolled far enough for it to enter their field of view — a result at the bottom of the page counts once the page was displayed.

So impressions measure eligibility rather than exposure, and a large impression count with almost no clicks frequently means your result was technically present and effectively invisible, rather than present and rejected. This matters because the two readings suggest opposite actions. If people saw your listing and did not choose it, the fix is the title and description. If they never saw it because you sit far down, rewriting the description changes nothing and the work is elsewhere. Distinguishing them means looking at position alongside impressions rather than at impressions alone: a low click-through rate at an average position of 4 is a snippet problem, and the same rate at an average position of 30 is not.

Click-through rate is clicks divided by impressions, and it is the most context-dependent number in the report. Aggregated across all queries, it is close to useless, because it blends two populations with entirely different intent. Someone searching your business name has already decided to reach you; they will click your result at a high rate almost regardless of what it says. Someone searching a generic term is choosing between providers and clicks at a far lower rate.

The practical error is reading a site-wide rate as a measure of how appealing your listings are. A business whose traffic is mostly people typing its name will show a healthy overall rate while performing poorly on every query that could bring a new customer, and the aggregate conceals exactly the problem worth solving. The fix is a filter: separate queries containing your business name from those that do not, and read the two groups apart. The unbranded group is where snippet work pays and where growth comes from. Watch the branded rate for one thing only — a sudden fall, which usually indicates something structural rather than a copywriting issue.

Amid the caveats, one pattern in this report is dependable and genuinely actionable: queries with substantial impressions and very few clicks, at a position good enough that people plausibly saw the listing. That combination identifies searches where demand exists, your relevance is established, and something about how you present or answer the query is losing the visitor. It is the closest thing to a free, unambiguous instruction available anywhere in search.

The second reliable use is discovering language. The query list shows the actual words people typed to reach you, which routinely differ from the words the business uses, and the mismatch is worth more than most keyword exercises. A third is diagnostic rather than strategic: a sharp change in impressions for a page, coinciding with a date you can identify, is a signal worth investigating even if the number itself is imprecise. None of these depends on average position being meaningful, which is why they hold. Notice what they have in common — each one compares your own site against itself over time or across queries, rather than treating an absolute figure as a measurement of your standing.

Several limits are structural rather than caveats. The data is sampled and low-frequency queries are omitted entirely, so the query list is not complete and an absence proves nothing. Queries are anonymised, meaning rare searches that could identify an individual are withheld — which particularly affects small sites with specific niches. There is a lag of a couple of days, so today's numbers are incomplete. And the report covers only queries where you already appeared, which means it can tell you nothing about demand you have never touched. That last limit is the one people forget: Search Console describes the visibility you have, never the visibility you do not.

It also cannot tell you anything about competitors. It is your property's data and contains no comparison to anyone else, and no free source provides that comparison — competitive visibility estimates are products sold by vendors, built from their own crawls, and they are estimates rather than measurements. So the honest summary is that Search Console is measurement of your own site, narrower than its metric names imply, and completely silent about everyone else. Used inside those bounds it is the best free resource available. Used as a ranking dashboard or a competitive tool, it produces confident conclusions that its own data does not support.