An agent once told me his website traffic was up 40 percent after a redesign. When I asked how many people had contacted him through the site, he did not know. That gap is the whole problem with real estate website analytics.
Why pageviews and bounce rate mislead agents
Pageviews are the first number every dashboard shows, which is why they get the most attention and deserve the least. A pageview counts an arrival. It says nothing about who arrived or what they wanted.
Traffic can rise for reasons that never turn into business. A post that gets shared in a Facebook group brings hundreds of visits from other agents. A page about mortgage rates attracts people from three provinces away. Both lift the number on the chart and neither lifts your income.
Bounce rate is worse, because it sounds like it measures disappointment. What it usually measures is a visitor who found what they needed on one page. Picture a seller who lands on your neighbourhood page, reads the recent sale prices, finds your phone number at the top, calls you, and closes the tab. In a bounce-style measure that visit counts as a failure. In your calendar it counts as a listing appointment.
The same logic breaks the "average time on page" number that agents often pair with it. A long visit can mean someone read carefully or it can mean they left the tab open while making lunch. Neither reading is safe on its own.
None of these numbers are useless. They are diagnostic. They help you understand a pattern after you already know something went right or wrong. They cannot tell you which it was.
The three numbers that answer the real question
Your website has one job: turn a stranger into a conversation. Three measures track that job directly.
Contacts, counted by source. Not visits. The number of people who called, filled in a form, or emailed, sorted by how they arrived. Organic search, Google Business Profile, social media, referral links, paid ads. This is the only number that connects the website to your income, and most agents do not have it because nobody writes the source down.
Which pages a visitor read before contacting you. The page someone contacts you from is often the contact page, which tells you nothing. The page before it is the one that did the persuading. On agent sites that is usually a neighbourhood page, a sold listings page, or the About page, and it is almost never the homepage. Knowing which one earns its keep tells you what to write more of.
Which search questions bring people in. Search Console shows you the actual words people typed before Google showed them your site. That is a free content brief written by your own market. If ten people a month reach you searching for a question you answer in one sentence, that question deserves its own page.
If you only ever record the first of these three, you are ahead of most agents.
Search Console and Analytics do different jobs
Agents mix these two tools up constantly, and the distinction is simple once you see it. One covers what happens before the click. The other covers what happens after.
Google Search Console is free and shows how your site behaves inside Google's search results. Google describes it as a service that helps you monitor, maintain, and troubleshoot your site's presence in Google Search results. It reports four things per search question: how many times your site appeared (impressions), how many times someone clicked, the resulting click rate, and your average position. It also tells you which of your pages Google has stored in its index and which it has skipped.
Nothing in Search Console covers what a visitor does on your site. It ends at the click.
Google Analytics starts at the click. It records which pages someone opened, how long they stayed, what they did next, and whether they completed an action you have marked as important, such as submitting a form. It cannot tell you what anyone searched for, because Google stopped passing that information to analytics tools years ago.
The practical split: use Search Console to decide what to write and what to fix in your page titles. Use Analytics to decide where your contact prompts go and which pages deserve more attention. Neither one replaces the other, and both are free.
Set up Search Console first if you have neither. It can show search data going back up to 16 months from the day you verify your site, so it hands you a history you never recorded. Analytics only records from the day you install it.
Why AI assistant traffic is hard to see
This is the part where honesty matters more than a tidy answer.
Google Analytics now has an AI Assistant channel in its default groupings, defined as visitors arriving from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok. Google also folds AI Overviews and AI Mode into the Organic Search channel, since those appear inside Google's own results.
So some AI traffic is visible. The trouble is what happens to the rest.
When a browser sends a visitor to your site, it normally passes along where the person came from. Analytics reads that and assigns a source. Links opened inside a phone app often carry no such information, and links a person copies and pastes carry none either. When the source is missing, the visit falls into Direct, which Google defines as arriving by a saved link or by typing your address.
That has a practical consequence for agents. Direct used to be a rough proxy for people who already knew your name. It is no longer safe to read it that way, because some unknown share of it is now people who asked an assistant a question and tapped a link in the answer.
Various tools and agencies publish percentages for how much AI traffic goes missing. We have not found a primary source we trust behind those figures, so we will not repeat them. What you can rely on is the mechanism: some AI referrals are labelled, some are invisible, and Direct absorbs the difference.
There is a low-tech fix that beats every analytics setting. Put one optional question on your contact form: how did you hear about me. A handful of people will type "ChatGPT" or "I asked an AI". That is the only evidence most agents will ever get, and it takes ten minutes to add. Our post on leads from AI search covers what those assistants read on your site before they send anyone.
Setting up so you can answer "did the rebuild work"
The day before a new site goes live is the last chance to record what the old one was doing. Miss it and the comparison is guesswork forever.
Capture four things and store them somewhere you will find again.
Contacts per month for the past six months, split by source. If you have never tracked source, count what you can reconstruct from your CRM and email, and note that the number is incomplete.
Total organic clicks and impressions from Search Console for the past six months. Export the file rather than taking a screenshot, because you will want to sort it later.
The list of search questions that produced your most clicks, and the average position for each.
The pages that received the most visits, with their contact counts if you have them.
After launch, resist judging anything for the first six to eight weeks. Google re-crawls a changed site gradually, and rankings move around while it does. The single exception is contacts: if form submissions stop dead on launch day, something is broken and you need to know within hours, not quarters. Test your own form from your phone the day the site goes live, and again a week later.
Then compare quarter to quarter, and compare against the same quarter a year earlier rather than the one just before it. Most real estate markets have a season, and a spring quarter beats a winter quarter whether or not anyone touched the website.
The quarterly review habit
Daily analytics checking is a way to feel busy. The numbers move around for reasons that have nothing to do with anything you did, and reacting to a bad Tuesday leads to changes that get reversed by Friday.
A quarterly review is the version that survives a busy season. Put 45 minutes in the calendar four times a year and answer five questions.
How many contacts came in, and from where? Compare against the same quarter last year.
Which pages preceded those contacts? Has that changed?
Which search questions gained clicks, and which lost them? A question that dropped is usually a page that has gone stale or a competitor who published something better.
Which pages got visits and produced nothing? Either they attract the wrong audience or they lack an obvious next step.
What did we add to the site since last quarter, and did it do anything? This question is how you stop paying for tools nobody uses. It pairs with the speed check in our post on website speed and mobile performance, since script stacks grow between reviews.
Write the answers in the same document every quarter. Four entries in, you have a record that tells you more than any dashboard, because it holds the context a chart never does.
What to do when the numbers say nothing is wrong
Sometimes the traffic is fine, the speed is fine, the search questions look right, and the phone still does not ring. That is a real pattern and it has its own diagnosis.
It usually means visitors arrive, read, and leave without an obvious reason to act. The page answered their question so completely that contacting you added nothing, or the next step was buried at the bottom, or the only option offered was a form asking for details they were not ready to give.
Analytics can point at this but cannot solve it. What it gives you is the list of pages where people arrive and nothing happens, which is the list worth rewriting first. Our post on traffic but no calls walks through that diagnosis in detail.
It also helps to remember what the visitor is doing while they read. In NAR's 2025 Profile of Home Buyers and Sellers, 70 percent of buyers used a mobile device or tablet during their search. A contact prompt that looks obvious on a desktop can sit three screens below the fold on a phone.
The takeaway
Pageviews and bounce rate are easy to read and hard to act on, which is why they dominate most agent dashboards and change nothing. Track contacts by source, the pages people read before contacting you, and the questions that bring them in. Use Search Console for what happens before the click and Analytics for what happens after. Accept that some AI traffic will stay invisible, and ask visitors directly instead. Then look at it once a quarter, write the answers down, and let the record build.
