Skip to content
SEO与数字营销

What Is KD in SEO? Read the Score Before You Budget for It

By Dizital Media· ·Dizital Media

Brand promotion: this article introduces Guangsuan's keyword difficulty guide and its GSR keyword ranking service. It is a vendor introduction, not an independent review.

Short answer: keyword difficulty (KD) is an SEO tool's estimate of how difficult it may be to reach the leading organic results for a query. It is not a Google score, and it is not your website's probability of ranking. KD is a way to order a candidate list inside one tool; it cannot tell you whether the searcher wants your page, whether you hold the evidence to answer the question, or whether the topic deserves the money. The output that survives a budget conversation is a page decision with reasons attached, not a list of green scores.

What a KD number actually measures

KD is a vendor calculation. Google does not publish a difficulty score, so each supplier builds its own model and names it differently. Guangsuan's guide to keyword difficulty collects the published definitions in one place — what the vendors say their metrics cover — and those definitions were checked in October 2026 rather than produced by a controlled comparison of paid accounts. Read them as vendor descriptions, not as a benchmark table.

MetricPublished scopeHow to use it
Ahrefs KDA 0–100 estimate of entering the top 10, based on the referring domains of the pages currently rankingA backlink-based estimate, not an assessment of content, intent or your own site
Semrush KD%A 0–100 estimate of the SEO effort needed to reach Google's organic top 10, using link, domain-authority and result-page inputsA general estimate that does not initially specify which domain you are researching for
Moz DifficultyA 1–100 score analysing the strength of the top 10 organic blue-link resultsUse Moz's own scale and method; its 40 is not another tool's 40
Mangools KWFinder KDA 0–100 difficulty metric using the strength of the pages on the first results pageKeep this general measure separate from the URL-specific Relative Keyword Difficulty

Even when tools use similar scales, they can analyse different inputs and weight them differently, so the same 40 does not describe the same competition in two products. The band names — easy, medium, hard — belong to the product that prints them; "easy" is not a promise that a new page will rank quickly. One unit is easy to misread as well: a referring domain is a website linking to a page, and one website can supply more than one backlink, so a vendor's referring-domain requirement is not a number of links you can purchase.

General difficulty, domain difficulty and page difficulty answer different questions

Some figures are not about the query alone. Semrush's Personal Keyword Difficulty estimates the difficulty for a named domain to reach the top 10, using topical relevance together with metrics for that domain and its competitors. Mangools describes Relative Keyword Difficulty for a named URL. A domain and a single page are different objects: those figures are not interchangeable with a general score, you cannot subtract one from another to produce a ranking advantage, and a percent sign is not a success probability. When you record one, write down the exact domain or URL you entered — a report produced for a different property, page or target market does not transfer just because the wording matches.

Why two tools disagree about the same keyword

Different models, different link databases, different result snapshots. The country or database, the language setting, the date the value was pulled, the refresh date the tool exposes, and the way close wordings are grouped all move the number. So a disagreement between two tools is not evidence that either score is unstable: instability would need comparable observations of the same metric under consistent conditions. It is also not a reason to average them. An average is a value neither vendor could reproduce, and it hides the settings that would let you reproduce their own number.

What to keep in the worksheet: the vendor, the product and the metric name, the value or missing status, the country or database, the language and other settings, the entered domain or URL where the metric is personalised, the collection date, and any result-refresh date the tool shows. Keep a missing value as missing rather than rewriting it as zero. Sort the working list with one metric, and use a second source only when it resolves a specific question. Keep the original values even when your editorial decision goes against the original score order.

Both numbers can be right about their own model. That is why the score is a sorting aid and the live results page is the evidence.

When a low score still means the page cannot move

A score describes the competition the tool sees for a query. It says nothing about the intent behind the query, the format the results reward, the state of your own site, or whether anyone buys behind the term. A low number can still describe a page you cannot move:

  • The intent does not match your page. A navigational query — someone looking for one brand's login page — can score low and still be wrong for a guide or product page you would publish.
  • The format is wrong. If the results hold comparison pages, specification pages or step-by-step tutorials, a general article is competing against a different job. Check whether you already have an appropriate URL to improve before planning a new page.
  • You cannot answer it yet. A specific query needs specific evidence: the measurements, documented procedure, compatibility notes, original test or relevant expertise the answer requires. An outline is not evidence.
  • The page cannot be reached, or cannot convert. A URL blocked from crawling, a directive that keeps it out of the index, a canonical pointing elsewhere because the material is duplicated or near-duplicated, or a page that gives a visitor nothing to act on will not move because somebody pushed links at it.

Two neighbouring columns in the same spreadsheet are not the same measure either. Search volume estimates describe demand; KD describes competition inside one tool's model. Google Ads uses the word "Competition" for something else again — the number of advertisers showing for a keyword relative to all keywords, under the selected location and Search Network. A high bid can prompt a commercial check; it does not establish your margin or prove that SEO work will pay. A low volume estimate does not prove that a specific customer question has no value.

A screening order that survives contact with real results

Teams hunting for "low competitive keywords" tend to start with the number and work backwards. The reverse order holds up better in a budget conversation:

  1. Demand first. Take an estimated monthly volume from one source, keep the country, the period and the rounding the tool displayed, and do not add a region to the countries inside it. A range is a range, not a count of buyers. When the market question spans more than one country, the site's guide to comparing keyword demand across countries before localising a page sets out the settings to hold constant when you pull those estimates.
  2. Order with one score, inside one tool, on one market setting. Treat that order as a queue for inspection, not as a ranking of opportunity.
  3. Read the live results. Open representative pages in the intended language and market. Record what each one answers, what format it uses, and what you would do differently. A forum thread can answer an experience-based question well, and an old publication date does not by itself make a page obsolete.
  4. Check what you already hold. List the evidence you can show today — specifications, steps you have run, photographs, records, compatibility notes — and name one gap you could close without inventing trials, client results or figures.
  5. Write the decision down. The query, the market, the tool and metric, the date, the value or missing status, the result URLs you inspected, the reader's task, the proposed page, the evidence required, and the next action: improve an existing page, gather material, commission an answer, or hold the topic.

That record is what makes a later review honest: a score changing does not prove your content improved, and a ranking change does not isolate the cause of that change.

On the page itself, Google's people-first content guidance asks whether the content provides original information, a substantial and complete description of the topic, and insight beyond the obvious — and it states there is no preferred word count. Difficulty is not one of its factors, and no difficulty score replaces the evidence the reader needs.

A worked example (teaching scenario)

The scenario below is invented for teaching. It contains no measured difficulty value, no search volume and no ranking result, and it describes no client.

Imagine a supplier of industrial storage cabinets weighing three candidate queries:

Candidate queryWhat the reader is trying to doPage that could serve itEvidence still needed
industrial storage cabinetsBrowse and compare available cabinetsA collection page with sizes, loads and filtersCurrent inventory, dimension drawings, load ratings
stackable storage cabinets for narrow aislesCheck whether a specific form factor fits the spaceA specification pageOutside dimensions, shelf clearance, stacking limits, aisle tolerances
how to measure an aisle for stackable cabinetsWork out whether the product fits before buyingA how-to with measurementsVerified measurement steps and drawings matching the shipped product

Suppose a tool later returns the lowest score for the broad query. That is a reason to inspect it, not a reason to publish an unrelated article. If the measuring question comes back with no score at all, record it as an open item rather than entering zero: an unanswered question may still deserve a page if you can supply the measurements.

Notice what set the order — the reader's task and the evidence available, not the colour of the number.

In-house or a ranking service: what to collect first

If you are deciding whether to handle a keyword in-house or to ask a provider, the decision does not start with the score. Collect these first; most of them are checkable before anyone sends an invoice.

  • Current position, with its conditions. The query, the landing URL, the country, the language, the device and the date. Averages reported over different windows describe different things, so fix the conditions before you compare two numbers.
  • Reachability of the target page. Check the exact URL: whether it can be crawled, whether anything keeps it out of the index, what the canonical points to, and whether it appears in the sitemap. A blocked URL, or one whose canonical sends visitors to a duplicate, will not move because a provider pushed links at it.
  • Whether the page can convert. What does the page offer, and what should happen after a click? If the answer is a form or a request, walk the route and confirm it works.
  • The gap between the results and your page. Record what each ranking page answers, and mark one gap you could close honestly.
  • Your own capacity. If the work needs technical access, original measurement, product documentation or a review cycle, name who owns it and how the material gets produced. A vendor score cannot estimate your project budget.

What to ask a provider before you pay

A provider should be able to answer these in writing:

  • Which metric, on which country and language setting, on which date?
  • What exactly is in scope, and what is explicitly out of scope?
  • How is the result checked, over what observation window, and who performs the check?
  • What triggers payment, and what happens if the target is missed?
  • Does the engagement depend on conditions you have not verified yet — indexability, page state, a specific existing position?

Where Guangsuan's GSR service fits this decision

GSR — Guangsuan Search Result Ranking System — is Guangsuan's keyword ranking service. Read its published entry condition first: the service opens on keywords that already hold a current natural ranking inside the top 20 for the agreed country, language and device. Guangsuan states that this threshold is its own business entry condition, not a qualification set by Google.

The published scope is off-site work around one key keyword and one target page, and it does not require backend access for content operations. Guangsuan is equally explicit that this does not mean the site foundation can be ignored: site and page status are checked before the engagement is confirmed, additional technical repair or content work is scoped separately rather than assumed to be included, a content gap can be linked to a separate article service, and a slow target page can be assessed separately for speed work. If a keyword has no ranking yet, or the site still has unresolved technical problems, Guangsuan describes integrated SEO as the broader work — technical work, content and link resources working together — rather than this ranking service.

Two further terms are published rather than implied. There is no upfront payment: settlement follows the ranking conditions written into the agreement, and the target position, checking method, observation window, payment trigger, annual start and maintenance are settled before the work begins. The published 30-to-45-day figure is a reference progression window, not a fixed promise — Guangsuan states that actual pace depends on competition, site foundation, page status and algorithm changes, and that Google decides search rankings, so no provider can guarantee a fixed position or immunity from updates.

The price structure is published on the same page rather than negotiated from scratch: GSR is charged as an annual amount for one keyword on one website, with tiers banded by monthly search volume, and the country, statistical date, tier boundary and keyword difficulty are confirmed at assessment and in the contract. Read the tier boundaries there rather than assuming that a page price list is the quote you have been given.

A first-page position is not traffic, enquiries or orders; that depends on the page, the offer and the market. The GSR entry condition, the scope, the steps from evaluation to acceptance and the annual per-keyword terms are set out on Guangsuan's GSR keyword ranking service page.

After the decision: review the outcome, not the score

After publishing, record what actually happened separately from the estimate: whether the page is accessible and indexable, whether it receives visibility, which queries bring it, and what enquiries or conversions follow. Compare the same queries and pages under comparable conditions, and revisit your assumptions when evidence accumulates rather than treating an arbitrary number of days as a pass-or-fail deadline.

Frequently asked questions

Is there a good KD score for a beginner?

There is no universal one. Start with relevant opportunities, use one tool's scale to organise them, then inspect the actual results and the work your site can deliver. A lower score inside one consistent tool makes a keyword worth checking earlier; it does not remove the need for those checks.

Two tools disagree about the same keyword. Which one is right?

Neither is necessarily wrong — they model competition differently and were read under different conditions. Keep both original values with vendor, metric, market and date, and use one metric to organise the list. A cross-tool disagreement is not by itself evidence that either score is unstable.

Does a high score mean I should drop the keyword?

Not on the score alone. Read the intent, the formats currently ranking and the evidence you hold, then choose to improve an existing page, gather material, commission an answer, or hold the topic.

Is Google Ads competition the same as KD?

No. That column describes the number of advertisers showing for a keyword relative to all keywords, under the selected location and Search Network. It is advertising competition, not organic difficulty.

What should I send when I ask a provider to assess a keyword?

The exact query, one target URL, the country, the language, the device, dated evidence of the current position, and the commercial reason the term matters. Agree in advance how the position will be checked, what the observation window is, what triggers payment, and what happens if the target is not reached.