PPC Competitor Analysis Without Mistaking Estimates for Facts

A PPC competitor analysis is a scoped review of other advertisers’ visible activity and your eligible auction comparisons. Its purpose is to form and test campaign hypotheses, not to reveal a competitor’s private account data. The strongest analysis answers a practical question such as, “Which message or keyword should we test next?”

For example, a rival domain appearing in Google Ads Auction Insights and a research tool estimating that the same domain advertises on a related keyword are different findings. Neither proves the rival’s budget, conversion rate, Quality Score, or complete keyword list.

This guide shows how to separate evidence types, compare equivalent scopes, preserve provenance, and move from research to a reviewable first-party test.

What a PPC competitor analysis can and cannot tell you

A useful analysis can help you identify advertisers that overlap with your eligible auctions, review observed ad messages and landing pages, discover estimated paid-keyword activity, and generate campaign ideas. It normally cannot expose a competitor’s private conversion rate, CTR, Quality Score, exact budget, or complete keyword inventory.

Keep these four evidence types distinct:

• First-party auction metric: a comparison reported in your own eligible Google Ads Auction Insights data.
• Direct observation: an ad or landing page reviewed with its query, market, device, and observation time recorded.
• Vendor estimate: a third-party database or model reporting paid keywords, ads, historical activity, traffic, or cost-related signals.
• Analyst inference: an interpretation such as “the offer appears aimed at price-sensitive buyers.” It is a hypothesis, not a fact.

Every finding should retain its source type, collection date, market, language, device, date range, and reporting grain. A competitor name beside a number is not enough context to compare periods or make a campaign decision.

Scope the analysis before looking for competitors

Begin with the decision. A branded-search defense review, a nonbrand keyword-gap review, and a promotion assessment require different queries and evidence. Set the comparable dimensions before collecting data:

• Country, location, and language
• Device and network
• Campaign type, such as Search, Shopping, or Performance Max
• Date range and comparison period
• Seed keywords, queries, or target domains

Separate a commercial competitor, which competes for the same customer decision, from an auction competitor, which appears in some of the same eligible auctions. A marketplace, directory, or adjacent service may overlap on a search without being a direct substitute.

Google Ads Auction Insights is useful for eligible shared auctions, but it is not a census of every advertiser in a market. Third-party research can broaden discovery, but its output must remain labeled as vendor research.

Save a scope record before collecting data. The following is an illustrative record shape, not a vendor template:

{
  "record_type": "analysis_scope",
  "business_question": "Find nonbrand search terms for a campaign review",
  "market_country": "US",
  "language": "en",
  "device": "mobile",
  "campaign_type": "Search",
  "date_start": "2026-09-01",
  "date_end": "2026-09-30",
  "seed_keywords": ["example service"],
  "scope_version": "v1"
}

If two reports differ materially in market, device, period, campaign type, or grain, do not treat them as a like-for-like comparison.

Read Auction Insights at the right grain

Google says Auction Insights is available for eligible Search, Shopping, and Performance Max campaigns. Search reports can be viewed at campaign, ad-group, or keyword level and segmented by time and device, subject to activity requirements. Google also notes that reporting is unavailable when the required activity threshold is not met and that insights are not shown when impression share is below 10%. A missing result may therefore reflect eligibility or reporting limits rather than an absent competitor.

For Search, the available measures include impression share, overlap rate, outranking share, position-above rate, top-of-page rate, and absolute-top-of-page rate:

• Overlap rate measures how often another advertiser received an impression when your ad also received one.
• Position-above rate measures how often another advertiser’s ad appeared higher when both ads appeared. This metric is specific to Search.
• Top-of-page and absolute-top-of-page rates describe placement relative to the top of the results page.
• Impression share and outranking share are comparative auction measures, not disclosures of another advertiser’s total market share or budget.

Record the entity, report grain, selected period, device, market, and campaign scope with every metric. Use a keyword-level report for a keyword question and a campaign-level report for campaign context. Store any rollup as a separate summary with its own aggregation definition. Auction Insights summarizes eligible activity; it is not an individual search-event log.

See Google’s Auction Insights documentation for current metric definitions, report levels, and eligibility details.

Build an evidence log, not a swipe file

Third-party advertising research can help discover estimated paid keywords, ads, historical activity, and landing-page URLs. Semrush documents competitor-domain, paid-keyword, ad-copy, historical-activity, and landing-page reports. It also explains that its database may not capture every ad because results vary by geography, scheduling, audience targeting, and keyword. Semrush describes its system as research based on its own database, not access to a competitor’s Google Ads account.

For each observation, capture the source tool and method, source URL, final URL where relevant, collection timestamp, market, language, device, query or keyword, original ad text, landing-page URL, and screenshot or export reference. Keep raw evidence separate from normalized themes such as offer, value proposition, or call to action.

Use Google Keyword Planner for historical search and bid-range context. Its historical metrics are distinct from forecasts, and search volumes can vary with location, language, network, seasonality, and date range. They do not verify that a particular competitor is currently bidding. Review Google’s Keyword Planner metrics guidance before selecting or comparing a metric.

Define the row grain before storing data. One ad-observation row should represent one observed ad for one query or keyword, market, device, source, and collection event. An Auction Insights metric belongs at its reported campaign, ad-group, or keyword grain in a different record type.

{
  "record_type": "ad_observation",
  "source_method": "direct_observation",
  "source_tool": "analyst_capture",
  "competitor_domain": "example.com",
  "market_country": "US",
  "language": "en",
  "device": "desktop",
  "query_or_keyword": "example service",
  "observed_at": "2026-10-09T14:30:00Z",
  "ad_text_hash": "sha256:illustrative",
  "raw_ad_text": "Illustrative headline and description",
  "landing_page_url": "https://example.com/landing-page",
  "collection_run_id": "run-2026-10-09-b",
  "analyst_inference": null
}

This proposed shape preserves the collection run and ad-text hash so two legitimate observations on the same day are not incorrectly collapsed. If a source supplies a stable record ID, preserve it as an additional identifier.

Compare landing pages without claiming conversion results

Review the page a visitor actually reaches. Follow redirects and record the final URL. Assess ad-to-page relevance, offer continuity, CTA clarity, form or checkout friction, pricing consistency, and visible trust evidence. These are analyst observations, not competitor performance metrics.

For technical checks, PageSpeed Insights provides performance diagnostics and suggestions. Record the requested URL, final URL, device mode, fetch time, lab results reference, and real-user data availability. Keep lab results separate from real-user data. Google’s documentation explains that PageSpeed Insights returns Lighthouse lab data and may include Chrome User Experience Report data, while current real-world data handling is moving toward CrUX APIs.

A speed result does not establish a competitor’s conversion rate or prove that a page element caused an outcome. Google identifies landing-page quality and relevance among auction considerations, but that does not show that a particular competitor page caused a specific auction result. Evaluate proposed page changes on your own site using analytics or controlled tests. See PageSpeed Insights documentation and Google’s auction-factor explanation.

Use a practical evidence-to-action workflow

The following examples are proposed operating patterns, not documented integrations or vendor templates. They show where a tool output ends and an analyst decision begins.

Trigger	AI job	Validation	Action or fallback
Auction Insights export at keyword or campaign grain	None for metric interpretation. Optional summary only from validated fields.	Check eligibility, activity limits, entity, dates, device, market, and report grain.	Store an auction observation. If scope is unclear, route to the PPC analyst and do not infer budget.
Observed ad or vendor ad export	Group supplied text into offer, CTA, value proposition, or intent themes.	Require exact evidence spans, confidence, source method, timestamp, and unchanged raw text.	Store normalized analysis separately. If evidence spans are missing, keep the raw observation and reject the classification.
Landing page selected for review	Summarize analyst notes into technical and message observations.	Check redirects, final URL, test conditions, lab versus real-user data, and page accessibility.	Link the page record to the ad observation. If inaccessible, record the failure and schedule a manual revisit.
Proposed bid, budget, keyword, targeting, or copy change	Draft a test hypothesis from linked evidence.	Confirm relevance, tracking, approval status, claim risk, and a measurable first-party outcome.	Assign a controlled test or close the finding. Never write directly to campaign settings from an inference alone.

Use deterministic rules for required fields, URL normalization, date parsing, source labels, text hashes, duplicate checks, legal terms, and grain validation. AI is better suited to bounded language tasks such as grouping supplied ad text. It should return fields such as detected_offer, detected_cta, evidence_spans, confidence, and human_review_required. It must not overwrite original text or convert an estimated spend signal into an exact budget.

Keep one immutable raw-observation record and a separate versioned interpretation record. A CRM contact or deal event is a separate operational record, not a substitute for either evidence type. Link any operational action back to the evidence IDs that support it.

For an ad observation, a proposed unique key should include competitor domain, platform, market, language, device, query, observed date, ad-text hash, landing-page URL, source tool, and collection run. For an aggregate, use a different key containing entity, platform, market, device, date range, metric, and aggregation definition. Enforce uniqueness with a database constraint and use a transactional upsert where supported. A lookup-then-create sequence alone is not race-safe when concurrent workers run.

Require human approval before using findings to change bids, budgets, targeting, or ad copy. Escalate trademark, pricing, guarantee, regulated, and comparative claims. Treat ad-copy similarity as an observation, not evidence of intentional brand diversion or infringement.

Choose tools by the evidence you need

• Google Ads Auction Insights: eligible same-auction comparisons at supported campaign, ad-group, and keyword grains.
• Google Keyword Planner: historical keyword metrics, bid-range context, and forecasts kept in separate fields.
• Semrush Advertising Research: vendor-reported competitor domains, paid keywords, ads, historical activity, and landing pages. Semrush documents CSV, Excel, and PDF exports for its Competitors report, but available fields and limits vary by plan.
• SpyFu: a possible source of vendor-reported keyword and ad-history research. Its current pricing page lists multiple plans and promotional rates, and API access and usage charges require separate plan and endpoint checks.
• PageSpeed Insights: performance diagnostics and suggestions, not conversion outcomes.

Do not quote a vendor price unless the exact edition and billing term are essential to the decision. Current documentation lists Semrush Advertising Toolkit Base at $99 per month and Pro at $220 per month, while SpyFu lists multiple plans and promotional rates. Ahrefs pricing varies by edition, with its current pricing page listing Lite at $129 per month and higher tiers above that. These figures are time-sensitive and should be checked immediately before purchase.

BuzzSumo and AdSpyder may support adjacent or cross-platform research, but their current PPC capabilities, pricing, coverage, exports, and API behavior were not sufficiently verified here to present as guaranteed workflow components. Use the smallest source set that answers the question. Auction Insights plus a small set of directly reviewed landing pages may be enough before purchasing broader research.

Establish ownership before moving findings between tools. A research database or controlled spreadsheet can hold observations, while a project record can hold the decision, owner, test, and outcome. If the handoff needs automation, define the fields and approval steps first through workflow automation. Decide which team owns operational records in its CRM systems before sending research there.

How often to repeat the analysis

There is no universal cadence. Tie review frequency to campaign volatility, spend exposure, seasonality, promotion cycles, and the decision being made. Establish a baseline before a significant campaign change, then repeat on a planned schedule or when a meaningful change in auction participation or observed messaging warrants review.

Compare equivalent scopes across periods. One newly observed ad, one temporary promotion, or one vendor database update is a signal to investigate, not proof of a lasting strategy change. Repeated evidence across matching markets, devices, dates, and queries is stronger, but it still does not reveal private performance data.

Every conclusion should be traceable to its source, scope, date, and evidence type. Every proposed campaign change should have a named owner, an approval status, and a first-party way to evaluate the result.