Fair publisher planning: giving independent sites a real chance
The short answer
Fair publisher planning means building site lists that judge independent, niche, local language and responsible sensitive topic publishers on the same evidence as large media groups. Relevance and quality decide the ranking, reach is shown as information but never used as a cutoff, missing data is never counted as failure, and the plan is audited slice by slice for bias.
Key takeaways
- Planning shortcuts, such as popularity cutoffs, keyword blocklists and domain endings, remove good independent publishers before anyone reads them.
- Public popularity data leaves out sites below an undisclosed threshold, so "no data" must never be read as "no audience".
- Responsible coverage of health, identity, crime and politics can be kept in a plan when suitability is judged by meaning, category by category.
- Bias in a plan can be measured: compare how each publisher slice survives from candidate pool to approved list, and why sites were removed.
- Fair planning keeps every quality check; it removes the checks that measure size instead of quality.
What does fair publisher planning mean?
Definition
- Fair publisher planning
- A way of building programmatic site lists in which each publisher is judged on its content, quality, suitability and selling path, and no group of publishers is removed by a rule that measures size, origin or wording instead of fit.
Four kinds of publisher can be lost to rules that measure size or wording instead of fit. Independent publishers that are not part of a large media group. Niche publishers with deep coverage of one subject and a modest audience. Local language publishers, such as French language sites in Quebec or Spanish language sites in the United States. And responsible sensitive topic publishers: health educators, community outlets and newsrooms that cover crime, conflict, identity or politics with care.
The stakes are real for local news. The Medill State of Local News 2026 report (opens in a new tab), released October 13, 2026, counts 238 US news desert counties, up 26 in one year, nearly 1,500 more counties with just one local news source remaining, and 47 million Americans with limited to no access to local news. It also tracked 75 local news startups in the past year, most of them digital only. Startups like these are small, new and independent, the profile a popularity cutoff can remove before anyone reads a page.
Why do standard planning rules penalize independent and local publishers?
Unfair outcomes need not be deliberate. They can come from shortcuts that were adopted to save review time and that happen to correlate with size. Each one is easy to replace with a check that measures what the buyer actually cares about.
| Shortcut | Who it can hurt | Fairer check |
|---|---|---|
| Minimum traffic or popularity cutoff | Niche, local and new publishers | Rank by relevance and quality; show reach beside each site |
| Keyword blocklists | News, health education and community outlets | Judge each page's meaning against the client's tolerance per category |
| Domain ending as a market signal | Sites based abroad with a real local audience | Look at audience by country, not the address |
| Missing data treated as a fail | Small sites absent from public datasets | Move unchecked sites to a separate unverified group with the reason |
| English only review | French and Spanish language publishers | Detect the language of each sampled article and plan for every language the brief allows |
| Lists built from brand recognition | Publishers the planner has not heard of | Start from article descriptions and seed sites, then read the candidates |
Platform settings can add to the effect. Google's content suitability page (opens in a new tab) describes its Limited inventory type as intended to exclude as much sensitive content as possible. Broad settings like that are reasonable for some brands, but they can remove sensitive and safe coverage along with harmful content, so they should be a deliberate client choice rather than a default.
Why should reach be shown in a plan but not used as a filter?
Reach matters to a buyer, and hiding it would be its own distortion. The problem is using reach as a gate. Public popularity data is coarse by design. The Chrome UX Report methodology (opens in a new tab) says origins that do not meet its popularity threshold are not included in the dataset at all, and Google does not disclose the threshold. Its popularity rank (opens in a new tab) is a log10 scale with half steps, such as the top 1,000, 5,000, 10,000, 50,000 and 100,000, and country datasets carry their own ranks.
Two consequences follow. A small publisher can be missing from the data while having a loyal, real readership. And a site in the top 100,000 for Canada cannot be told apart from other sites in the same band using that data alone. A plan that drops every site without a band, or keeps only the top bands, removes independent publishers on evidence that cannot support the decision.
How can a plan keep responsible coverage of sensitive topics?
A sexual health clinic's guide to testing, a local newspaper's court report, a disability advocacy magazine and an LGBTQ+ community site all contain words that keyword rules were built to catch. None of them is the kind of content those rules were meant to stop. The difference lies in how a page treats its subject, which only a reading of the page can show.
Score each sensitive category separately, with graded levels of risk, and set the client's tolerance per category rather than one overall safe score. The industry's shared vocabulary for content is the IAB Tech Lab Content Taxonomy (opens in a new tab). The World Federation of Advertisers discontinued GARM (opens in a new tab) in August 2024. Its Brand Safety Floor and Suitability Framework is a legacy reference, not an operating standard or certification. The brand suitability guide covers tiers and tolerances in depth.
How do you measure bias in a media plan?
Bias in a plan shows up as one group of publishers surviving the process at a lower rate than others, for reasons unrelated to fit. It can be measured with a simple slice audit. Slices describe publishers and pages, never people: ownership type, size band, country of base, language and whether the site covers sensitive but safe topics.
A five step slice audit for any site list
- Tag the candidate pool. Before filtering, label every candidate with its slices: independent or group owned, size band, base country, main language and sensitive topic coverage.
- Measure survival by slice. For each slice, divide the sites on the approved list by the candidates in the pool. A slice that survives at a much lower rate needs an explanation.
- Read the removal reasons. Count the reasons sites in each slice were removed. Reasons tied to quality or suitability are legitimate; reasons tied to missing data, popularity or domain ending point to bias.
- Check the unchecked group. Compare the share of each slice sent to the unverified group. Small sites piling up there signal data gaps, not risk.
- Fix the rule, not the result. When a slice is served worse, change the rule that caused it and rerun, rather than adding a few sites back by hand.
| Measure | How to calculate it | Warning sign |
|---|---|---|
| Survival rate | Approved sites in a slice divided by candidates in that slice | Independent or local language slices survive at a lower rate than group owned sites |
| Removal reason mix | Share of removals in a slice by reason | Removals driven by missing data or popularity rather than quality |
| Unchecked share | Unverified sites divided by candidates in the slice | One slice is mostly unchecked while others are fully reviewed |
| Rank position | Median position of the slice's approved sites | Equally relevant sites from one slice sit consistently lower |
How does PlacementGraph keep planning fair to independent publishers?
What should a fair site list checklist include?
Fair planning checklist before approval
- No site was removed only for low popularity, a missing popularity band or its domain ending.
- Sensitive categories were judged by page meaning against the client's tolerance, not by a keyword list.
- Sites that could not be checked sit in a separate group with the reason, not in the excluded group.
- Every language the brief allows was searched and reviewed, including Canadian French where relevant.
- Survival rates were compared for independent, niche, local language and sensitive topic slices.
- Any slice served worse has a written cause and a rule change, not a manual patch.
- The plan shows each site's reach band as information beside its reasons.
What does a fairness review change on a real plan?
Fictional example
A fictional Canadian health insurer's wellness campaign
A fictional regional health insurer briefs a display plan beside wellness and preventive health content in New Brunswick and Manitoba, in English and French. The first draft list, built with a popularity cutoff and a stock keyword list, holds almost only national media brands.
The planner runs a slice audit. French language sites survived at a fraction of the English rate, mostly removed for having no popularity band. Health education sites on menstrual and sexual health were all removed by keyword matches, although none crossed the client's tolerance when the pages were read.
The team replaces the cutoff with a visible reach band, swaps the keyword list for category tolerances judged by meaning and reruns the search. The approved list now includes an Acadian community health magazine, two independent nutrition publishers and a sexual health education site, each with its reasons and reach shown.
Guides in this hub
Every guide in the Fair publisher planning hub, each written to answer one question in full.
- GuideHow can brands advertise beside addiction recovery content?How to keep addiction recovery and harm reduction coverage in a media plan: which brands fit, how to read the language, and how to write the policy clause.
- ChecklistWhat should an AI audit trail record for each planning decision?A checklist of what to record with each automated planning decision, from model version and inputs to reviewer and date, so a media plan can be audited later.
- ExplainerHow can AI site recommendations skew a plan toward large publishers?How AI site recommendations can skew a media plan toward large publishers, the NIST bias types behind it, and tests and questions to audit an automated plan.
- GlossaryAlgorithmic fairness definitions, applied to publishersPlain definitions of parity, calibration, equal error rates and slice based evaluation, applied to how a model includes or excludes publishers, never people.
- GuideHow should a media plan treat a bilingual website with uneven coverage?How to plan ads on bilingual websites when one language section is thin: judge each section on its own, read language signals, and show sections in the plan.
- ExplainerDo brand safety tools treat some communities unfairly?Research on classifiers that flag identity terms and dialect as unsafe, what it means for community publishers, and a method buyers can use to test a tool.
- GuideHow do you include Canadian independent publishers?Who Canada's independent digital publishers are, how the Online News Act and Meta's news block affected them, what the LJI funds and how to plan them in.
- GuideHow do you keep disability content in a media plan fairly?Why disability news, advocacy and accessibility content gets wrongly blocked, how to tell it from injury content, and how to describe it in a policy.
- FrameworkHow should a brand set and track a diverse media spend target?How brands set a spend target for diverse owned, local and independent media, decide which publishers count, track delivery and report progress honestly.
- GuideHow do you include diverse owned media in a plan with definitions you can verify?Plan diverse owned media advertising with definitions you can verify: ownership tests, certifiers in the US and Canada, claims to avoid, and a checklist.
- GuideHow can brands advertise beside domestic violence awareness?Why domestic violence awareness and survivor support content gets blocked as crime, how to tell it from violent content, and how to write a clear tolerance.
- ExplainerWhat makes an AI site recommendation explainable?What an explainable AI site recommendation should show, from supporting pages to its weakest point, and five tests a planner can run to see if it is real.
- ExplainerWhich governments require advertising in community media, and how is it documented?Which governments require ad spend with community and ethnic media, what New York City's Local Law 83 requires and how buyers document spend for each agency.
- ExplainerHow is Government of Canada advertising governed, and what does it mean for publisher choice?Government of Canada advertising rules: the 2025 policy and directive, Official Languages Act duties, who coordinates buys and what it means for publishers.
- GuideShould brands advertise beside health news?How brands can advertise beside health journalism on outbreaks, research and public health, why it was blocked in 2020, and how to write a disease news policy.
- GuideHow do you plan in-language media into a US campaign?Why US in-language community media in Chinese, Korean, Vietnamese and other languages drops out of plans, and how to verify, review and report it.
- GuideHow do you include Indigenous media in a media plan, respectfully?How to include Indigenous owned and Indigenous language media in US and Canadian plans: who they are, how to verify ownership, where language detection fails.
- GuideHow can advertisers keep investigative journalism in reach without blunt rules?Why investigative outlets covering abuse, crime and corruption lose ad revenue to blunt rules, and the tolerance settings and reviews that keep them on a plan.
- ExplainerWhat is language variant bias, and how do buyers test for it?How classifiers and filters can serve Canadian French, Mexican Spanish or regional English worse than standard forms, and how buyers test quality per variant.
- GuideHow do you keep LGBTQ+ content in a media plan responsibly?Why LGBTQ+ news and lifestyle pages get wrongly blocked, what published research shows, and a review method and policy clause that keep them in a media plan.
- GuideHow do you plan advertising on local news sites in the US and Canada?How to plan advertising on local news sites in the US and Canada: what counts as local news, how much is left, how outlets sell, and a market checklist.
- GuideShould brands advertise beside mental health content?Keep mental health coverage and support content in a media plan: how suicide reporting guidelines shape suitability, what needs extra care and what to exclude.
- ExplainerHow does keyword blocking affect news publishers?How news keyword blocking works, what sourced studies found about its effect on news publishers, where Google applies keyword exclusions, and a blocklist audit.
- ExplainerWhat are news deserts, and what can advertisers do about them?What a news desert is in US and Canadian research, how advertising money shaped local news closures, and practical steps buyers can take in their media plans.
- GuideHow should a nonprofit plan programmatic advertising on a small budget?How nonprofits plan programmatic campaigns: mission aligned publishers, suitability for sensitive causes, small budgets, election rules and reports to ask for.
- FrameworkWhen should a planner override an AI site recommendation, and how is the call recorded?When a planner should override an automated site recommendation, what evidence justifies it, how to record the reason, and how override patterns reveal bias.
- GuideHow should brands plan media for Pride Month and after it?Plan Pride Month media on LGBTQ+ publishers that stay on the plan after June: why pull outs happen, suitability settings for Pride news and how to report it.
- GuideHow do public service announcements run in programmatic?How public service announcements run in programmatic media, what donated inventory is, which publishers suit public health messages and how to report PSAs.
- FrameworkHow should a buyer run a fair publisher appeal?How a buyer designs a fair appeal for publishers excluded for suitability, MFA risk or quality: grounds, evidence, an independent reviewer, timelines and fixes.
- FrameworkHow do you measure publisher concentration in a media plan?Measure how much of a media plan goes to one owner or network, find sites that share an owner, set a concentration limit and report it to clients clearly.
- GuideHow do you plan beside religious content without blanket exclusion?Plan beside faith coverage without blanket exclusion: separate religious publishing from hateful content, write a policy clause and review the creative.
- FrameworkResponsible media investment: a framework for brandsA framework for brands whose spend should support quality journalism and diverse publishers while meeting suitability and performance goals, with measures.
- GuideWhy do rural publishers fall out of programmatic plans, and how do you bring them back?Why rural publishers fall out of programmatic plans, how missing popularity data hides them, and how to verify rural media and report rural reach honestly.
- ExplainerWhy do satire sites get flagged, and how should advertisers review them?Why satire and humor sites get flagged as misinformation or unsafe, how to tell satire from fake news, which brands fit, and how to review a satire site fairly.
- GuideHow do you keep sexual health education in a plan and explicit content out?How to keep responsible sexual health education in a media plan while excluding explicit content, with a decision table by content type and a review method.
- ExplainerWhy do small websites have no CrUX data, and what should a plan do?Why small and niche websites can have no Chrome UX Report record, why that gap is not an audience of zero, and how to show missing data honestly in a plan.
- GuideHow do you include small YouTube creators in a plan fairly?Why subscriber minimums drop strong small YouTube channels, which evidence can stand in for size, and how to test small creators in a plan without buying junk.
- GuideAre student newspapers good placements, and how do you plan them?How to find, verify and plan independent student newspapers and campus media for youth brands, from ads.txt on university subdomains to teen reader rules.
- ExplainerHow accurate are third party website traffic estimates, and what should a plan use instead?How accurate third party website traffic estimates are, why they fail small and niche sites, how they bias media plans, and better public evidence to use.
- GuideHow do you keep women's health content in a media plan?Why menstrual, menopause and reproductive health pages get blocked, which brands fit beside them, how to treat reproductive news, and policy wording to use.
Frequently asked questions
Does fair publisher planning mean lowering quality standards?
No. Independent and niche sites still face the same checks for editorial quality, made-for-advertising risk, suitability and selling path. Fair planning removes the rules that measure size, origin or wording instead of quality. A small site that fails a quality check stays off the plan, with the reason recorded.
Are small independent sites more likely to be made-for-advertising sites?
Size alone is not evidence either way. Made-for-advertising sites are defined by their business model: bought traffic, heavy and refreshing ad loads, generic content. Judge each site on those signals, and treat a small audience as a fact to show, not as a risk signal.
How do I define an independent publisher for a slice audit?
Use a definition you can apply consistently and write it down, for example a site not owned by a company that runs several large publications. Shared ad or analytics codes across several domains can point to common ownership. Keep the definition about the publisher, never about its readers.
Should sites sold only through resellers be removed from a fair plan?
Not on that basis alone. Small publishers can rely on intermediaries to sell their space. Check that the site's ads.txt file and the exchange's sellers.json entry agree, note that the path is a reseller, and let the trading desk decide whether that path suits the deal.
Can a buyer prove a plan is unbiased?
No plan can be proven free of bias, but it can be measured and documented. Record survival rates and removal reasons by slice for each version of the list. That record shows the client which rules shaped the plan and lets the team improve the rules over time.
Sources
- Local news deserts skyrocket as newspaper closures escalate, even as startups emerge, Medill study shows (opens in a new tab)Medill School of Journalism, Northwestern University, October 13, 2026
- The State of Local News 2026 (opens in a new tab)Medill Local News Initiative, Northwestern University
- CrUX methodology (opens in a new tab)Google Chrome for Developers
- CrUX metrics (opens in a new tab)Google Chrome for Developers
- About content suitability (opens in a new tab)Google Ads Help
- IAB Tech Lab Content Taxonomy (opens in a new tab)IAB Tech Lab
- WFA discontinues GARM (opens in a new tab)World Federation of Advertisers, August 9, 2024
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.