How to Track Keyword Rankings Accurately
Learning how to track keyword rankings accurately comes down to fixing every variable that changes a result: the location, device and language of the search, how deep you look, and how you count positions when ads, AI Overviews and other features sit above the classic links. Check the same keywords under the same settings on a fixed schedule, store every result, and judge the trend rather than any single day.
Most ranking arguments between an SEO lead, a client and a tool happen because two people are measuring different things. This guide defines what a rank is, covers each setting that moves it, explains why Search Console disagrees with your rank tracker, and ends with a method for choosing keywords and reading the numbers.
What a keyword ranking actually measures
A keyword ranking is the position at which a URL from your site appears in the results for a query, under one specific set of search conditions. The number only means something together with those conditions. “Position 4” without a location, device, language and date is an anecdote, not a measurement.
Tools and Google also count positions differently, which is the first source of confusion. There are three common counting methods:
| Counting method | What it counts | Where you see it |
|---|---|---|
| Organic rank | Only classic organic results, numbered 1, 2, 3 in order | The default view in most rank trackers |
| Absolute rank | Every result element on the page in order, including image packs, video carousels and AI Overviews | Trackers that offer an “absolute” or “all results” view |
| Search Console position | The topmost position of any link to your site, where a feature block counts as one position and ads count as none | Google Search Console |
Google documents its rules in the Search Console help page on impressions, position and clicks. Positions run top to bottom down the main column, a carousel occupies one position that all its items share, an AI Overview occupies a single position shared by every link inside it, and ads do not take a position at all.
One results page counted three ways
Picture a mobile results page for a software query: two ads, then an AI Overview that cites your page among four sources, then a competitor’s organic result, an image pack, a second competitor, a video carousel, and finally your page as a classic organic result.
| Element, top to bottom | Organic rank | Absolute rank | Search Console position |
|---|---|---|---|
| Two ads | none | none | none |
| AI Overview citing you and three others | none | 1 | 1 |
| Competitor A | 1 | 2 | 2 |
| Image pack | none | 3 | 3 |
| Competitor B | 2 | 4 | 4 |
| Video carousel | none | 5 | 5 |
| Your page | 3 | 6 | 6 |
For this one search, a tracker in organic mode reports you at 3, a tracker in absolute mode reports 6, and Search Console records 1, because it keeps the topmost link to your site and the AI Overview citation sits above your organic listing. All three numbers are correct. They answer different questions: how do I compare with other web pages, how far down the screen am I, and where was the highest link to my site?
The practical rule is to pick one counting method for reporting, write it on the report, and record SERP features as separate flags (cited in the AI Overview, owns the featured snippet, appears in the video carousel) instead of folding them into the rank.
The search conditions that move a ranking
Google returns different results to different people, and most of the difference is not personal history. Google’s own help page on search location says it always estimates a general area from your IP address, and may also use precise device location, saved home or work addresses, and past activity when you are signed in. Each of those is a variable your tracking has to pin down.
| Variable | Why it moves rankings | How to control it |
|---|---|---|
| Location | Results are localized from IP address, device location and saved places | Set a fixed location per keyword: country for national terms, city or postal code for local intent |
| Device | Mobile and desktop pages use different layouts and show different features | Track mobile and desktop as separate datasets |
| Language | Interface and query language change which pages and features appear | Set the language explicitly, especially in multilingual markets |
| Signed-in activity | Past searches and visits can nudge results | Use clean, logged-out checks from a neutral environment |
| Time | Results and features change from day to day | Check at a consistent time on a consistent schedule |
| Surface | Web results, Maps, AI Overviews and AI Mode are different products | Track each surface as its own dataset |
Location deserves the most care. For a broad informational query, the top ten is usually similar across a country. For anything with local intent, such as “accountant” or “emergency plumber”, the leading results can change completely a few miles away. A rank checked from your office tells you where you rank in your office, nothing more.
Device is the second biggest factor. Mobile pages stack more features above the first organic link, so the same organic rank can sit much lower on a phone screen. Track both devices and weight your reporting toward the one that brings most of your traffic.
How deep and how often to track
Depth used to be cheap. Until September 2025, a tool could load 100 results in one request by adding num=100 to a Google search URL. Around 10 to 12 September 2025 the parameter stopped working, and Google told Search Engine Land that the parameter is not something it formally supports. Seeing positions 1 to 100 now takes ten page requests instead of one, so a typical rank tracker tool now defaults to the top 20 or top 50, or charges more for deeper checks.
The change also rewrote Search Console data. Tyler Gargula’s analysis of 319 Search Console properties found that 87.7% saw impressions fall and 77.6% lost unique ranking queries, while average position jumped for many sites. The common explanation is that bots loading 100-result pages had been generating impressions at deep positions; when those disappeared, averages improved without any real ranking change. Annotate mid-September 2025 on every year-over-year chart, or you will misread it as a win.
For depth, ask what decision the data supports. Most ranking decisions concern the first two pages. Tracking deeper is useful mainly as an early signal that a new page is indexed and climbing, so a top-50 or top-100 check once a week is usually enough for that job.
Frequency follows the same logic:
- Daily: revenue keywords, during site migrations, redesigns, launches and known Google updates.
- Weekly: the bulk of a keyword set, where you care about direction rather than day-to-day noise.
- Monthly: long-tail and experimental keywords you mainly want to report on.
Search Console average position vs a tracked rank
Search Console reports the topmost position of your site for each impression, averaged across every search in the period: every location, device, language and query variant combined. A position is recorded only when an impression happens. A keyword rank tracker reports one simulated search under conditions you chose. Expect the two to differ.
Here is a worked example. A page ranks 3rd for a keyword in the United States and 14th in the United Kingdom. In one week it earns 100 impressions in the US and 40 in the UK.
Average position = (100 × 3 + 40 × 14) ÷ (100 + 40) = 860 ÷ 140 = 6.1
A tracker set to the US reports 3. Search Console reports 6.1. Neither is wrong; one measures a fixed market, the other blends every market that saw the page. Add the AI Overview effect from the earlier table, where a citation can lift your recorded position to 1, and the gap can run in either direction.
| Question | Use Search Console | Use a rank tracker |
|---|---|---|
| Which queries actually bring impressions and clicks? | Yes | No |
| Where do I rank in Chicago on mobile today? | No | Yes |
| Where do competitors rank for my keywords? | No | Yes |
| Did a SERP feature appear or disappear? | Partly | Yes |
| What is my click-through rate? | Yes | No |
Use both. Search Console tells you what real searchers saw; a tracker tells you what changed and why, under controlled conditions.
AI Overviews and AI Mode on the results page
AI features now sit on top of many results pages, and they change what “ranking” means. In Search Console, an AI Overview is one position, and every link inside it shares that position. John Mueller has said a position for AI Overviews is hard to report “in a way that makes it useful”. In AI Mode, each element gets its own position, and follow-up questions count as new queries.
Since June 2026, Google has offered generative AI performance reports in Search Console, rolled out to all sites by the end of August 2026. They show impressions, pages, countries and devices for AI Overviews and AI Mode, but no clicks, and their numbers are a subset of the Web totals, so never add them together. That is useful context, but it does not tell you which competitor was cited instead of you or how often you appear for a given prompt.
For tracking, treat the AI layer as its own measurement. For each keyword, record whether an AI Overview appeared, whether your page was cited, and which domains were. An organic number one does not guarantee a citation, because AI search engines choose which sources to cite by their own retrieval logic. Measuring presence in those answers is the core of AI visibility. Asheard, for example, tracks classic Google positions in its rank tracking and, separately, the mentions and citations a brand earns in Google AI Overviews, AI Mode and eight other AI engines.
How to choose keywords to track
Part of knowing how to track keyword rankings is knowing which ones deserve tracking at all. Track the keywords that tie to business outcomes first, and volume second. A keyword set built only from high-volume terms produces reports that look busy and explain nothing.
| Keyword group | Example | Why track it |
|---|---|---|
| Money keywords | “payroll software for restaurants” | Closest to revenue; track daily |
| Category terms | “payroll software” | Shows market position against the main competitors |
| Comparison and alternative queries | “[competitor] alternatives” | High intent, often contested by review sites |
| Problem and question queries | “how to calculate overtime pay” | Feeds the funnel and often triggers AI Overviews |
| Striking-distance queries | Queries at positions 8 to 20 in Search Console | The fastest wins once content is improved |
| Brand queries | “[brand] pricing”, “[brand] login” | Detects hijacked or lost brand results |
Map each keyword to the one URL meant to rank for it. When a different URL from your site starts ranking instead, that is a cannibalization or intent signal you would miss by watching the number alone. Tag every keyword by cluster, intent, funnel stage and target location, so you can filter reports by what the business cares about.
As a rule of thumb, most small and mid-sized sites get a clear picture from 50 to 200 well-chosen keywords. Add more only when each new group answers a question the current set cannot.
How to track keyword ranking trends without fooling yourself
Single-day movements are mostly noise. Read the shape of the data instead.
- Use position buckets, not one average. Count keywords in positions 1 to 3, 4 to 10, 11 to 20, 21 to 50 and not ranking. A move from bucket to bucket is easier to explain than a fractional average.
- Know how unranked keywords are handled. If nine keywords sit at position 5 and one drops out of the top 100, a tool that counts the missing keyword as 101 reports an average of (9 × 5 + 101) ÷ 10 = 14.6. One keyword made the whole set look three times worse.
- Smooth daily data. A 7-day median shows direction without reacting to every reshuffle.
- Annotate events. Content updates, migrations, Google updates and the September 2025 Search Console shift all belong on the chart.
- Compare with competitors on the same keywords. If everyone dropped two absolute positions overnight, a new feature probably appeared above the organic results; nobody lost ground in organic terms.
- Pair rankings with clicks. Rankings up while clicks stay flat usually means features above you are absorbing attention, which is where GEO and SEO start to diverge.
Checklist for how to track keyword rankings accurately
- Choose one counting method (organic or absolute) and state it on every report.
- Set a fixed country, city or postal code for each keyword based on its intent.
- Track mobile and desktop separately.
- Set the search language explicitly.
- Assign a target URL to each keyword.
- Record AI Overview presence and citations as separate fields.
- Pick a frequency per keyword group: daily, weekly or monthly.
- Keep the full history, and annotate the September 2025 change and your own site changes.
- Review bucket counts and competitor movement before reacting to any single keyword.
Frequently asked questions
What is the most accurate way to track keyword rankings?
Use a rank tracker that checks each keyword from a fixed location, device and language on a regular schedule, logged out and without personal history. Store every result, report one counting method, and record SERP features such as AI Overviews separately. Then cross-check the trend against Search Console, which shows what real searchers saw across all locations and devices.
Why does my rank tracker show a different position than Search Console?
Search Console averages the topmost position of your site across every impression, location and device in the period, and a citation inside an AI Overview can count as your top position. A rank tracker reports one simulated search under fixed settings. A page ranking 3rd in one country and 14th in another will show both numbers correctly.
Do AI Overviews count as a ranking position?
In Search Console, yes. An AI Overview occupies one position, and every link inside it shares that position, so being cited can give you position 1 even if your organic listing is lower. Most rank trackers report organic rank without it, which is why AI Overview citations should be tracked as a separate field.
How many keywords should I track?
Track enough to cover the outcomes you care about, not every term with volume. For most small and mid-sized sites, 50 to 200 keywords grouped by cluster, intent and target URL give a clear picture. Add a new group only when it answers a question your current keyword set cannot.
Did Google’s num=100 change affect rank tracking?
Yes. Since September 2025, Google no longer returns 100 results per request, so checking the top 100 takes ten requests and many tools now track fewer positions by default. Search Console impressions also dropped and average position improved for most sites, a reporting shift rather than a real change in rankings.