How do you find similar websites, starting from one seed site, and check them?
Part of Lookalike websites
The short answer
To find similar websites from a seed site, describe the seed's typical articles in plain sentences, gather candidates from searches, the seed's links and its writers' other outlets, then sample several recent articles from each candidate. Keep only sites where most samples match, and run publisher, seller, audience and quality checks before anything joins the plan.
Key takeaways
- Start from sentences describing the seed's typical articles, not from the seed's domain name or its category label.
- Candidates come from several routes: descriptive searches, the seed's outbound links, its contributors' bylines and the sources it cites.
- A candidate is similar when most of its recent articles match the seed's descriptions, judged on a sample pulled from its sitemap.
- Every similar website still has to pass publisher, ads.txt, sellers.json, audience and made-for-advertising checks before it is planned.
What should a planner write down before searching for similar websites?
Searching straight from a domain name can return that domain's rivals for search traffic, which may be large, general sites. A better starting point is a short profile of the seed that a colleague could use without ever opening it. Writing the profile takes a few minutes and turns a vague request into criteria that can be tested on every candidate.
| Field | What to write | Why it matters later |
|---|---|---|
| Seed URL and section | The domain, or the section path when only part of the site fits | Stops the search drifting toward the seed's other sections |
| Three article sentences | One plain sentence each for three typical recent articles | These sentences become the search phrases and the match test |
| Depth and voice | Expert test, hobbyist diary, local news or how-to guide | Separates a close match from a site on the same subject at another depth |
| Languages allowed | The languages the brief accepts for this theme | Decides whether translated matches count |
| Target countries | Where the campaign will run | Sets the audience check for each candidate |
| Hard exclusions | Subjects or site types the client never wants | Removes candidates before anyone spends time reading them |
Where do candidate websites come from when expanding from a seed?
No single route finds every similar site, so combine several. Google's current search operators documentation (opens in a new tab) lists only filetype:, imagesize:, site: and src:, with no related: operator. Expanding from a seed is now a manual research task unless a tool does the matching.
| Route | How it works | Strength | Weakness |
|---|---|---|---|
| Descriptive searches | Paste each article sentence into a search engine, adding quotation marks around a distinctive phrase | Finds sites by subject even when they never mention the seed | Search ranking can place large, well linked sites ahead of small specialists |
| The seed's outbound links | Read the seed's resource pages, blogroll and links inside articles | Editors may link to publications they respect | Links also point to stores, sponsors and friends' sites |
| Contributor bylines | Search the names of the seed's regular writers to see where else they publish | Writers with a specialty may also write for similar outlets | Slow, and some writers also produce paid content for brands |
| Cited sources | Follow the studies, newsletters and specialist outlets the seed cites | Surfaces niche and trade publications | Cited sources can be institutions, not ad supported publishers |
| Section searches | Use site: with a topic word on a large site to find one fitting section | Finds matching sections inside big publishers | Google says site: results are not guaranteed to include every indexed URL |
Google's site: documentation (opens in a new tab) warns that the operator does not necessarily return all indexed URLs and that results without a search term are relatively random, so treat it as a way to spot a section, not to count articles. A minus sign before a word, as Google Search Help (opens in a new tab) describes, removes results containing it, which helps drop retail pages that crowd out publishers.
How do you test whether a candidate is really similar to the seed?
A homepage shows what a site promotes this week, not what it publishes. The test should run on a fresh sample of articles, chosen so that the planner cannot pick the flattering ones. A site's XML sitemap can help: the sitemaps protocol (opens in a new tab) defines a lastmod tag as the date of last modification of the page, which lets a planner pull the most recently updated articles instead of browsing.
Six steps to judge one candidate against the seed
- Pull five recent articles. Open the candidate's sitemap, which may be listed in its robots.txt file, and take the five newest article URLs by lastmod, skipping tag pages and product pages.
- Write one sentence per article. Describe each article the same way the seed profile describes the seed's articles.
- Score each article. Mark it as a match when it could swap places with one of the seed's three sentences without the client noticing.
- Keep the share, not the best hit. Record how many of the five matched. Three or more out of five suggests a candidate worth checking further.
- Note the depth and voice. A matching subject at a very different depth, such as news briefs against long expert guides, goes on a separate list.
- Save the evidence. Keep the URLs of the matching articles beside the site, so a reviewer or the client can verify the call later.
Which checks does a similar website need before it joins a plan?
Similarity only says a site covers the right subjects. It says nothing about whether the site can be bought, whether real readers in the target country visit it, or whether it was built mainly to resell ad impressions. These checks use public files and public data, so they can be done before any money moves.
Checks for each similar website
- The site is a publisher with its own articles, not a store, a brand site or a directory of links.
- An ads.txt file sits at the root of the domain, where the IAB Tech Lab specification (opens in a new tab) says publishers should post it, and it lists at least one DIRECT account.
- Each DIRECT seller ID in ads.txt appears in that exchange's sellers.json (opens in a new tab) file, with a seller type that matches the claim.
- The OWNERDOMAIN value in ads.txt, where present, was compared across candidates, since a shared owner suggests one network.
- The site appears in the Chrome UX Report (opens in a new tab) data for the target country, or its absence is noted as missing data rather than as no audience.
- The language of the sampled articles is one the brief allows.
- The pages passed a made-for-advertising review covering ad density, refreshing slots and bought traffic, as set out in the MFA guide.
The OWNERDOMAIN check is quick. The ads.txt 1.1 specification describes it as the business domain of the entity that owns the site. When four candidates in a fresh list declare the same owner, the planner is looking at one company's portfolio and should decide whether one entry, or none, belongs on the plan. The supply path guide covers the seller checks in more depth.
What does finding similar websites look like on a real brief?
Fictional example
A senior dog food brand in Washington and Oregon
Tovenmark Pet Foods, a fictional maker of food for older dogs, sells through independent pet stores in Washington and Oregon. Its agency's only seed is a veterinarian run blog about caring for aging dogs. The planner writes three sentences: a vet on protein needs after age eight, a guide to joint friendly exercise for older dogs, and an owner's account of adjusting a home for a dog with poor eyesight.
Descriptive searches return two large pet portals and several specialist blogs. The seed's resource page links to a canine rehabilitation clinic's newsletter, which carries no advertising and drops out, and to a dog sports magazine. One of the seed's writers also contributes to a regional pets and family magazine in Portland.
Sampling settles the calls. One pet portal matches on one of five recent articles, mostly covering puppies and product roundups, so it leaves the list. The dog sports magazine matches on two of five and goes to an adjacent list. Two specialist blogs declare the same OWNERDOMAIN and share a layout, so they are reviewed together as one network. The final list records the matching article URLs beside each remaining site.
How does PlacementGraph find similar websites from a seed?
What should the finished list of similar websites record?
A list of domains alone forces the next person to redo the research. A few extra columns let a buyer, a client or an auditor see each decision and check it again months later, when sites change owners or topics.
| Column | What it holds |
|---|---|
| Site | The domain, or the section path when only one section fits |
| Seed matched | Which seed and which of its article sentences the site resembles |
| Match share | How many sampled articles matched, out of how many read |
| Evidence | The URLs of two matching articles |
| Seller path | DIRECT or RESELLER, and the exchanges confirmed in sellers.json |
| Audience note | Present in Chrome UX Report data for the target country, or missing |
| Status and date | Core list, adjacent list or left out, with the reason and the date checked |
Keeping the date matters because every check here can change. A blog can be sold, a seller account can be closed and a site's subject can wander. With dated evidence, the next review starts from what changed instead of from scratch. For lists that must hold up in Canada and the United States at once, the audience targeting guide explains how country data should be read.
Frequently asked questions
Can a search engine alone find similar websites?
It can find candidates, but not judge them. Search results rank pages for a query, which can favor large and well linked sites, and they do not show seller files, audience by country or ad density. Use searches to build the candidate pool, then sample articles and run the checks on each site yourself.
Should the seed site itself go on the final list?
Yes, provided it passes the same checks as the candidates. A client may assume a favorite site is safe, yet it may sell only through resellers or carry a heavy ad load. Run the seed through the publisher, seller, audience and quality checks, and record the results beside it like any other site.
What if a similar website has no ads.txt file?
Treat it as a gap to resolve, not an automatic rejection. An ads.txt file is how a publisher declares its authorized sellers, so ask the publisher how it sells its space or whether it takes direct deals. Note the missing file and the date you checked, and keep the site off programmatic deals until the selling path is clear.
Do similar websites have to be based in the same country as the seed?
No. What counts is whether readers in the campaign's target country visit the site and whether its articles are in an allowed language. A publisher based elsewhere can have a strong local readership, and a local domain ending says little about who reads a site. Check country level audience data instead of the address.
Sources
- Overview of Google search operators (opens in a new tab)Google Search Central
- Google site: search operator (opens in a new tab)Google Search Central
- Refine Google searches (opens in a new tab)Google Search Help
- Sitemaps XML format (opens in a new tab)sitemaps.org
- ads.txt (Authorized Digital Sellers) (opens in a new tab)IAB Tech Lab
- ads.txt Specification Version 1.1 (opens in a new tab)IAB Tech Lab
- sellers.json (opens in a new tab)IAB Tech Lab
- CrUX methodology (opens in a new tab)Google Chrome for Developers
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.