AdWords competitor keyword tools: what each one actually returns

An AdWords competitor keyword tool gives you one of four things: modelled keyword estimates from SpyFu, Semrush or Ahrefs, first-party data from Google's Keyword Planner, Transparency Center and Auction Insights, a queryable API layer, or a free scraper. Buy by the evidence you need to defend, not by database size. Most paid-search teams need one modelled tool plus the auction data they already own.

The phrase is a legacy one. AdWords became Google Ads in 2018, and the tools that grew up around it still sell the same promise: type a domain, get the keywords it bids on. What changed is that the promise is now easy to check, and once you check it, most of the market looks the same underneath.

What an AdWords competitor keyword tool can and cannot see

No public source lists the keywords another advertiser bids on. Google does not publish competitor keyword sets, bids or budgets. Every tool that claims to show them is inferring, and the inference rests on one of three things: an ad that appeared for a query, a landing page that ranks for a term, or a panel-based volume estimate attached to either.

The observation behind a paid-keyword row is a sample. Scrapers run a fixed query list, at a fixed frequency, from a fixed set of locations, and record which advertiser's ad appeared. A keyword the competitor bids on but never matches inside that sample never enters the database. Coverage is a function of crawl breadth and crawl freshness, not of truth.

So the honest framing for any of these tools: they show you a sample of a competitor's footprint, translated into volumes by a model you cannot inspect. That is genuinely useful for prioritisation. It is not the same as seeing the account.

Where the keyword numbers actually come from

Three inputs feed almost every volume figure in the market: Google's own planner bands, clickstream panels, and scrape counts from search results. Each has a different error profile, and the price of a licence rarely tells you which one you are buying.

Input 1: Keyword Planner bands

Keyword Planner returns an average monthly figure over a trailing 12 months. For accounts without meaningful spend it collapses volume into wide bands instead of a precise number, so a keyword sitting in a 1K-10K band tells you the order of magnitude and nothing about whether the fifth digit is a 1 or a 9.

Input 2: Clickstream panels

Panels estimate volume from a sample of browsing behaviour and scale it up to the population. They are internally consistent, which makes them good for ranking two keywords against each other, and weak on absolute size. Two vendors can use overlapping panels and still publish numbers 30% apart after their own smoothing.

Input 3: Scrape counts

Some tools derive volume from how often a term shows up in search results or autocomplete. That measures supply, not demand, and it is the cheapest input to collect, which is why it appears in budget tiers. Treat any volume column you cannot trace to a named input as decoration.

The four routes, and what each one costs in 2026

List prices move, so treat the figures below as a snapshot taken in October 2026 from vendor pricing pages and third-party price trackers. The shape of the market is far more stable than the numbers.

Route 1: Modelled databases

• SpyFu: Basic at $39 a month billed monthly or $33 a month billed annually, with a free tier capped at 10 results per search instead of 5.
• Semrush: the entry SEO plan lists at roughly $140 a month on monthly billing, and Keyword Gap sits inside that plan rather than behind an upgrade.
• Ahrefs: Starter at $29, Lite at $129, Standard at $249 and Advanced at $449 on monthly billing. Competitor Analysis and the Ahrefs API and MCP access unlock at Lite, not Starter.

Route 2: Google's first-party surfaces

Keyword Planner is free inside any Google Ads account and returns bands plus top-of-page bid ranges. Auction Insights is free and shows impression share overlap against the domains you actually compete with. The Transparency Center is free and shows live and past creative by advertiser. None of the three lists a competitor's keyword set, and all three are measured rather than modelled.

Google's transparency surfaces deserve their own treatment, because they are the only first-party record of what an advertiser actually ran. Our breakdown of what the Google Ads Transparency Center API really exposes covers the per-ad field list and why it is thinner than the marketing suggests.

Route 3: Ad intelligence APIs and MCP servers

API layers sit above the scrapers and return structured records you can query, diff and schedule. This is the route that changes the shape of the work: instead of a dashboard a human remembers to open, you get JSON an agent compares week over week and reports when a domain changes behaviour. Our PPC competitor analysis framework sets out where that comparison belongs inside a campaign-review loop.

Route 4: Free and manual

Incognito searches, the Transparency Center and a spreadsheet. Slow, unglamorous and surprisingly good at one job: confirming that a keyword a paid tool showed you is being bid on this week.

How accurate are third-party keyword volumes?

Less accurate than the interfaces imply, and there is a vendor-published number worth quoting rather than guessing at. Ahrefs' help centre, in an article dated 3 September 2026, reports its own search volume estimates as roughly accurate for 60% of the keywords studied when compared against impressions in Google Search Console, and puts Google Keyword Planner at roughly accurate in 45% of cases.

"Ahrefs' Search Volume estimates were roughly accurate for 60% of the studied keywords compared to Impressions from Google Search Console... Google Keyword Planner's Search Volume estimates were roughly accurate in 45% of cases." Source: Ahrefs Help Centre, 3 September 2026.

Read that carefully. The better performer in the comparison is directionally right about three keywords in five. Build a budget model on a single volume column and you are compounding a 40% miss across every downstream calculation.

What survives the noise is relative ordering. When two tools both put keyword X well above keyword Y, that ranking is usually real even when both absolute numbers are inflated. Use volume to sort a list, not to compute revenue.

How do you get competitor keywords without paying for a tool?

Four free moves, ordered by how fast each produces usable evidence.

• Auction Insights by domain over 30 days. This tells you which competitors actually overlap your impressions, and it is the only free signal that is measured rather than estimated.
• Search your top 20 terms twice a week and log which domains show ads. Ad copy changes are the earliest public signal of a bid or offer change.
• The Transparency Center, filtered to one advertiser and to Search. You see the ads and their start dates, not the keywords behind them.
• Their landing pages. A competitor publishing pages around a topic is bidding on that topic, and page velocity is cheaper to read than any keyword list.

None of this replaces a database, but it does replace the first month of one. Our walkthrough on finding which keywords competitors bid on in Google Ads covers the ordering, the match-type traps and the gap-analysis loop in more detail.

Why keyword lists matter less after AI Max

Google moved AI Max for Search campaigns out of beta in a post dated 15 April 2026 and updated on 11 June 2026, pushing the Dynamic Search Ads sunset and auto-upgrade to February 2027 after advertiser feedback. The direction did not change: keyword lists become signals that guide matching, not gates that stop it.

That has a direct consequence for tool selection. If your own account matches against intent rather than exact strings, then a competitor's keyword list is a weaker proxy for their targeting than it was in 2019. Their ad copy, their landing pages and their search-term reports carry more of the signal.

Three things still hold regardless of match type: the domains you overlap in Auction Insights, the pace at which they ship creative, and the terms their pages rank for. None of those depend on a keyword set anyone published.

How to test a keyword tool in 25 minutes before you pay

Run this on a trial, not on a sales demo. The test is built to fail a tool cheaply and early.

• Pick 10 keywords whose performance you already know from your own search-term report.
• Enter the competitor domain in the tool and record which of the 10 appear, and at what volume.
• Pull the same 10 from Keyword Planner and record the band for each.
• Pull impressions for the matching queries from Search Console or from the search-term report.
• Score two numbers: recall, the share of the 10 the tool surfaced, and skew, its volume divided by your impressions.

A tool that finds 4 of 10 and runs consistently 40% low is still usable for ranking. A tool that finds 9 of 10 with wild variance across the set will break every model you put on top of it.

SELECT search_term_view.search_term, metrics.impressions, metrics.clicks FROM search_term_view WHERE segments.date DURING LAST_30_DAYS ORDER BY metrics.impressions DESC LIMIT 100

That query is the ground-truth half of the test. It costs nothing, it is first-party, and it is the only column in the comparison you can defend in a review meeting.

The same discipline applies once you turn volume into money. Read how to estimate a competitor's ad budget from public ad data before you multiply anything by a CPC, because the error bar moves faster than the estimate.

Which tool fits which buyer

Four buyer shapes, four defaults. None of them is two databases plus a dashboard.

Solo founder under $100 a month

SpyFu Basic or Ahrefs Starter plus Auction Insights. You are buying one thing: a faster read on which themes a competitor invests in. Adding a second database doubles the cost and halves the attention.

Small agency, three to ten clients

Semrush or Ahrefs Lite plus a fixed report template. The cost that hurts here is not the licence, it is an analyst re-pulling the same views every week. An API route earns its price at roughly five recurring clients.

In-house performance team

An API or MCP layer above a database, feeding the warehouse you already have. You own the first-party half of the picture; the tool exists to add the competitor half on a schedule.

Anyone who has to defend a claim

Auction Insights first, always. Estimated volume is a hypothesis and impression share overlap is a measurement. When the two disagree, the measurement wins and the hypothesis gets rewritten.

What to do next

Before buying anything, export 30 days of Auction Insights for your top campaign and list the domains holding more than 5% impression share overlap. That list, not a keyword database, is the set of competitors actually worth researching.

Then run the 25-minute test on the two tools you were already considering. Most teams find the cheaper one wins on the specific keywords they care about, and the expensive one wins on breadth they will never use.

Frequently asked questions