What this workflow saves you from
Do not compare a GST calculator site with a generic investment portal or a software store with a video editor just because a few words overlap.
Do not create pages because a competitor repeats words such as “track”, “right”, “studio” or “content”. Prefer repeated user-task phrases.
A CAPTCHA, 403 or unreadable sample is an evidence failure—not permission to invent competitor findings.
Public HTML does not reveal private traffic, conversions, all rankings or backlinks. Use real connected data when those metrics matter.
Step 1 — establish your own site identity
Start with your own URL. Review the detected subject, site/business type and publishing format. If the site is mixed—such as technology news, a video-led travel blog, or a software marketplace with tutorials—the label should preserve that context instead of forcing the site into one unrelated bucket. Add a short audience/context hint only when the public pages genuinely cover multiple subjects.
Step 2 — choose a competitor by task, audience and model
A candidate is not a confirmed competitor merely because it appears in discovery. Prefer the site that serves a similar visitor task, market and operating model. A global category leader can be useful as a quality benchmark, but its topic gaps may be irrelevant to your business.
| Signal | Strong direct competitor | Adjacent benchmark |
|---|---|---|
| Subject | Same specific niche/task | Only broad category overlap |
| Audience | Similar visitor problem | Different audience/use case |
| Business/site type | Similar store/tool/publisher model | Different model |
| Sampled topics | Several meaningful shared phrases/terms | Zero or very weak overlap |
| Action | Research gaps after validation | Use mainly for UX/technical benchmarking |
Step 3 — check sample completeness before reading gaps
Look at both sample cards. A five-page interactive target is still a bounded sample, not a full-site crawl. If the competitor is blocked or only one page is readable, treat the comparison as incomplete. Deep saved-project monitoring can expand public-HTML coverage, while JavaScript-rendered evidence requires a render bridge.
Step 4 — use repeated phrases, not loose words
GapForge v186 extracts 2–3 word phrases from fetched titles, H1s and headings and counts support by distinct sampled pages. A topic can enter the BUILD shortlist only when it is supported by at least two competitor pages and the relevance gate is open.
Weak evidence: “track”, “right”, “studio”, “content”, or a phrase seen on only one page.
Step 5 — open the source pages
Do not act on the phrase card alone. Open its source URLs. Confirm that the phrase represents a real page/task rather than navigation text, a footer, a one-off promotion or a misleading context. The evidence links are part of the result on purpose: audit the audit.
Step 6 — validate demand with real data
Repeated competitor coverage proves only that the competitor publishes that topic. It does not prove search volume or ranking opportunity. Use your own Search Console for queries/impressions and a licensed keyword provider if you need market demand. Build only when the topic fits your audience and does not cannibalize an existing page.
Step 7 — measure after the change
Record what you changed, the target page, the date and the expected user task. Then monitor impressions, clicks, CTR, positions and conversions. Competitor research should produce a measurable experiment, not an endless content-production queue.
What to do when CAPTCHA/WAF appears
Do not try to bypass it. If SocialKit itself shows a verification page, complete the verification and review your own WAF/Cloudflare rules for the GapForge same-origin API endpoints. If the competitor blocks the crawler, choose another readable competitor or use an authorized data source. GapForge reports challenge HTML as a security/fetch limitation rather than a valid result.
Seven-question pre-publish checklist
- Is this competitor actually relevant to the same visitor task?
- Did the competitor sample fetch enough readable pages?
- Is the gap a repeated phrase rather than a generic word?
- Can I open at least two source pages that support it?
- Does my audience genuinely need this topic?
- Does Search Console/licensed demand data support the decision?
- Can I create a clearly better, original page without copying?