How to Increase Website Traffic: What Real Data Shows

Most advice on how to increase website traffic starts with "publish more", but three months of Search Console and GA4 data from two real sites point to a different order of work.

Key takeaways

• Split traffic into brand, non-brand, page type and country before you change anything. Both sites below looked like they were moving one way and were actually moving the other.
• Off-topic posts can earn tens of thousands of impressions and almost no visitors. When one site lost them, its daily impressions fell by 90%, yet within four weeks its daily clicks were back to their spring level.
• Crawl waste (staging subdomains, tag feeds, file downloads) keeps Google away from the pages you want indexed, including your About and author pages.
• The best visitors came from places most traffic plans ignore: the map pack, branded searches and ChatGPT, whose visitors stayed over five minutes.
• Count leads before you chase visits. One site grew sessions by 72% and recorded zero key events.

The two sites and the data#

Both are real sites, anonymised here. We compared the last three months (28 July to 27 October 2026) with the three months before in Search Console and GA4, then checked backlink data and the live results pages for their main queries.

• Site A is a digital marketing agency in Delhi: about 600 indexed pages, service pages plus a blog that mixes Google news with "list of sites" posts.
• Site B is a young SEO blog: about 110 indexed pages, mostly Google and AI-search news, plus downloadable templates.

Look at the first two rows. Site A lost half its impressions and gained clicks. Site B gained impressions and lost a third of its clicks. Judge either site by the headline chart and you fix the wrong thing.

Step 1: Split your traffic before you try to grow it#

Totals hide the cause. Before planning anything, break the comparison into four cuts in Search Console (Performance → Search results → date filter → Compare → "Compare last 3 months to previous period"):

1. Brand vs non-brand. Add a Query filter → Custom (regex) → "Doesn't match regex" with your brand name and its misspellings, for example brandname|brand name|brnd. Switch it to "Matches regex" for the brand-only view.
2. Pages. Sort by the clicks difference, then by the impressions difference, and group winners and losers by page type: service pages, guides, news, tools.
3. Countries. A loss in a country you don't serve is a different problem from a loss at home.
4. Devices. Desktop-heavy impressions at low positions often come from research queries that rarely turn into visits.

Here is what the split showed.

Site A's loss was almost all foreign, off-topic impressions. United States impressions fell from 32.7k to 12.0k, yet US clicks only went from 13 to 12. India, where the agency's clients are, rose from 128 to 185 clicks, and its average position there improved from 52.8 to 40.6. Desktop carried 89% of all impressions.

Site B's loss was almost all navigational. The homepage fell from 58 to 23 clicks, the brand name from 32 to 19, and "site:" searches for the domain disappeared. That usually means the owner and early readers stopped checking the site through Google, not that rankings fell. Its non-brand pages gained impressions.

Step 2: Cut the traffic that was never going to convert#

Site A's blog had four posts aimed at broad, worldwide research queries: an explainer on SEO in digital marketing and lists of website archive, blog submission and directory submission sites. Together they earned 52,000 impressions and 59 clicks in the earlier three months, roughly one click per 900 impressions, at average positions in the 40s. None of them had anything to do with what the agency sells.

The chart at the top of this guide is Site A's daily data. Impressions averaged about 1,400 a day in the weeks before 20 August 2026, then fell below 100 and stayed there. That was the third day of the August 2026 spam update. There was no manual action, and timing alone doesn't prove the cause, but the posts that dropped were thin "list of sites" pages in a link-building niche that Google's spam policies watch closely.

Clicks fell too, from about 4 a day to about 1 a day for three and a half weeks. From mid-September they settled at about 2.3 a day, slightly above the spring average of 2.1, on a tenth of the impressions. Over the quarter, the four posts lost about 34,700 impressions but only 37 clicks. They looked big in the chart and were worth very little.

What to do with posts like these:

• Off-topic and thin: noindex or remove them. A Delhi agency gains nothing from ranking 45th in the US for "web archive".
• On-topic but thin: merge several weak posts into one strong one and 301-redirect the old URLs to it.
• On-topic and close to page one: improve them. They are your fastest wins.

Repeat this check every quarter: filter Pages for more than 1,000 impressions and fewer than 5 clicks, and ask of each page whether it could ever lead to a customer or a returning reader.

Step 3: Stop Google wasting crawls on the wrong URLs#

Site A's Page indexing report showed 614 pages indexed and 1,335 not indexed. The biggest groups:

• 937 "Not found (404)": 860 on a test subdomain and 49 on a second staging subdomain. Both still served live pages, their robots.txt allowed crawling, and one listed its own sitemap. In six months, 164 staging URLs earned impressions, competing with the real pages.
• 228 "Crawled, currently not indexed", including dozens of tag archives, their RSS feeds and several service pages.

Site B had a smaller version of the same problem. Its 25 crawled-but-not-indexed URLs were mostly Word and Excel files, WordPress scripts and feeds, and its 33 "Discovered, currently not indexed" URLs included the About page and the author page. Google knew the trust pages existed and hadn't fetched them.

Fix them in this order:

1. Put staging sites behind a password. HTTP authentication keeps every crawler out. A robots.txt block alone can leave URLs in the index, and a noindex tag only works if Google can crawl the page and see it. Then request removal of the staging host in Search Console and delete its sitemap.
2. Noindex tag archives and stop linking to feeds unless they rank for something useful.
3. Link to the pages you want crawled. Add an author box that links to the author and About pages on every post, put both in the main menu and the XML sitemap, and link from the homepage to your best guides. Internal links are how a small site tells Google what matters.
4. Re-check in two to four weeks. Inspect a few fixed URLs with the URL Inspection tool, then watch the "Indexed" line in the Page indexing report.

Our robots.txt and sitemap checker shows in seconds what a host's robots.txt allows and which sitemaps it lists.

Step 4: For local businesses, win the map pack first#

Site A's money keyword is "digital marketing agency in" plus its area. The results page for it showed, in order:

1. a map pack with three local businesses;
2. a business directory listing;
3. small agencies' own pages;
4. "Top 10 agencies in" the area, written by other sites.

Site A ranked around 84th. No amount of blog content changes that layout. These do:

• A complete Google Business Profile for one real address, with the same name, address and phone number on the site, in LocalBusiness schema and in directories. Site A's schema named one neighbourhood while several service pages targeted another where it has no office. Target the area you can prove.
• Reviews. A search for the agency's name plus "reviews" appeared this quarter with 224 impressions. People check before they call, so ask every satisfied client for a review.
• The directories and listicles that already rank. If a "top 10 agencies" page holds position 9, getting listed on it is faster than outranking it.

Google names relevance, distance and prominence as the main local ranking factors in its tips to improve your local ranking.

Step 5: Put the asset on the page, not in a download#

Site B earns close to 1,000 impressions a quarter for "seo strategy template", at an average position of 43. Its templates exist, but only as .docx and .xlsx downloads, and Search Console lists those files as crawled but not indexed.

The page ranking first shows its template on the page, and so do most of the smaller sites on page one: a structured outline you can read and copy, often with a preview image, with the download as an extra. An image pack sits in second place, so a clear screenshot of the template can earn a second listing.

If you offer templates, checklists or calculators:

• Build the template as HTML on the page, with headings that match what people search for ("SEO strategy template", "SEO plan template").
• Add a preview image with descriptive alt text.
• Keep the file download, but as a bonus, not the content itself.

Step 6: Build evergreen hubs alongside the news#

News posts get a burst of fresh rankings and then slide. On Site B:

A young site can't hold news rankings once publishers with tens of thousands of linking domains cover the same story. What held up on Site B were practical guides, such as getting pages indexed in Brave Search (still at positions 15 to 20), and one update report that stayed around position 8.

Keep covering news, but pair it with pages that keep their value:

• Keep one evergreen hub per recurring topic, for example every Google core update, or how each AI assistant finds its sources, and update it whenever something changes.
• Link each news post to its hub, and the hub to the latest news.
• Change the hub's updated date only when you change its substance.

Step 7: Earn links with numbers nobody else has#

Site B has 9 referring domains, 5 of them dofollow. Nearly all of them point to posts with original numbers, such as an estimate of how many searches an AI assistant runs, cited with anchors like "35% of those prompts trigger web search". None point to the posts that summarised other people's news.

Site A has 303 referring domains, but the profile is weaker than the count suggests:

• 2,276 links are "managed by" footer credits on client sites.
• 125 domains link with the same exact-match anchor for its main service keyword.
• 83 low-cost domains carry an identical, auto-generated anchor.

Whoever built them, links like these are a risk, not an asset.

What works:

• Publish one piece of original data a month, such as a survey, a count or a test with screenshots, and send it to the writers who cover that topic.
• Make footer credits on client sites brand-only and nofollow.
• Audit for automated link patterns. If links look like a paid or automated network and you can't get them removed, a disavow file is the documented last resort.

Step 8: Treat AI assistants as a traffic channel#

Site A's best-engaged visitors didn't come from Google.

AI assistant sessions went from 4 to 100 in three months. The numbers are still small, but these visitors arrive with a specific question, and they read.

To see this channel and grow it:

• In GA4, open Reports → Acquisition → Traffic acquisition. If the channel list has an "AI Assistant" row, start there. If not, switch the dimension to Session source and look for chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com.
• Check which pages they land on, and write more like them: a direct answer near the top, original numbers and clear headings.
• Make sure robots.txt doesn't block the AI search crawlers you want, such as OpenAI's OAI-SearchBot.

Step 9: Measure what the traffic does#

Site A grew sessions by 72%, users by 61% and engaged sessions by 78%, and recorded zero key events in both periods. No contact-form submissions, phone taps or WhatsApp clicks were tracked, so nobody can say whether the extra visits brought a single lead.

The same report hid two more problems. Direct traffic made up 73% of sessions with a 28% engagement rate, a sign of bots and internal visits. A referral from "localhost" showed that a developer's test visits were being counted.

Before your next traffic push:

1. Mark form submissions and phone and WhatsApp clicks as key events in GA4.
2. Define internal traffic (Admin → Data streams → your stream → Configure tag settings → Define internal traffic) and activate the internal-traffic data filter.
3. Tag every campaign link with UTM parameters so it doesn't land in Direct. Our UTM builder does it in one step.

A 30-day plan#

Then run the three-month comparison again. If clicks from the pages and countries that matter are up, and key events are being recorded, your traffic is growing in the way that counts.