Discussion starter: a publisher and community SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Indexing
- News SEO
Syndication partners outrank your original stories. Canonical requests, noindex agreements, or delayed syndication: what worked?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Of the three, only a noindex on the partner’s copy, kept in place, reliably stops it competing with your original in Google. Google’s documentation moved that way in May 2023. A cross-domain canonical request is a weaker hint that Google may not follow when the two pages look different. Delayed syndication addresses a different part of the problem: it gives your original the first part of a story’s life to itself, but it does not stop the partner outranking you once its copy is live.
The main qualification is that this is a commercial negotiation before it is a technical fix. A partner that noindexes your stories gives up search traffic from them, so the realistic outcome rests on your leverage and on what the deal is worth compared with the search, News and Discover traffic you are losing. A large publisher supplying a portal with content it wants can ask for noindex. A smaller publisher that needs the portal’s reach will probably end up with delay, excerpt or canonical terms, and should price the deal with that loss in mind.
What the public evidence shows
Public evidence is thin, and no controlled public test comparing the three approaches exists that I can point to. The most detailed public case study I can point to is Glenn Gabe’s analysis of about 3,000 syndicated news articles (published August 2023, updated September 2025). It checked one publisher’s recent URLs with the URL Inspection and Search Console APIs across Search, Top Stories, the News tab and Google News, and found Discover too personalised to track well. Three points matter here:
- Almost all of the original articles were indexed, and the study found no cases of Google selecting the partner’s URL as the canonical.
- Even so, many originals had little or no performance data. Partner copies at times held Top Stories or News tab positions while the original ranked lower or was missing, and which version won looked inconsistent from one surface to the next.
- Gabe summarises Google’s May 2023 change as a full recommendation that partners noindex a publisher’s content when the publisher does not want to compete with them in Search. The reasoning is that the original and the partner’s page can often differ once the whole page is counted, including boilerplate and supplementary content. He is also candid that asking partners to noindex will probably not go well.
The first point changes the diagnosis. In that dataset the problem was not canonicalisation, where Google folds your page into the partner’s. Both pages were indexed as separate documents and competed for the same slots. A fix aimed at canonical selection works on the wrong layer when Google already treats the two pages as distinct. This is one publisher’s observational data: a reason to check your own URLs, not a rule.
Why a partner’s copy can beat the original
Think of each news surface as a shop window with room for one copy of each story. Both copies arrive. A canonical tag on the partner’s copy is a note saying “the original is over there”, which the shopkeeper may or may not follow, particularly when the copies come in very different covers. A noindex means the partner’s copy never goes on that shop’s shelves, although the partner can still hand it to its own readers. Delay means your copy arrives first and has the window while the story is new, but the partner’s copy can still take the slot once it turns up.
Technically, the partner’s page is a different URL on a different site with different surrounding content, and it is evaluated as its own document. Large partners bring things your page may lack: sections that are crawled very frequently, internal links from high-traffic hubs, strong site-level signals, and sometimes a rewritten headline that matches the query better than yours. Publishing first helps, but it does not guarantee that your version is the one shown for a given query on a given surface.
How the options compare
Option Effect in Google Cost to the partner Where it falls short Evidence behind it Partner noindex ( robotsmeta)Partner copy leaves Google’s index, so it cannot appear on any Google Search surface Loses search traffic to that copy from Google and other engines that honour the tag Partners resist; templates change, so it needs auditing Google’s 2023 recommendation, as summarised in the case study Google-only noindex ( googlebotmeta)Same effect in Google Keeps visibility in Bing and services built on other indexes You still compete with the copy outside Google A negotiating compromise, not a documented recommendation Cross-domain canonical to your URL A hint Google may use if it treats the pages as duplicates Low: the partner keeps its page and the hint may be ignored Pages with heavy boilerplate may not be treated as duplicates Google stopped recommending it for this purpose in 2023 Delayed syndication Your original can be crawled and shown before a competing copy exists Loses freshness on breaking stories Partner can outrank you once live; little value for evergreen pieces No platform documentation; timing logic Excerpt plus link to the full story Partner’s page is a less complete answer to the query Less content to monetise Can still rank for the headline; some partners refuse No platform documentation; reasoning The Google-only noindex matters because it may be the only version a partner will accept. Its cost to you is visibility outside Google: the copy can stay in Bing’s index and surface in services built on other search providers; OpenAI, for example, documents that ChatGPT search sometimes partners with third-party search providers, alongside its own OAI-SearchBot crawler.
Delay is worth more for breaking news than for analysis and features. For breaking news, a short window covers the hours when interest in the story is highest. An explainer earns traffic for weeks, and the partner’s copy competes for all of it once the window closes. Match the window to the content type rather than applying one number to everything.
What I would check before asking a partner for anything
- Name the surface. Web results, Top Stories, the News tab, Google News and Discover behave differently. Search Console separates Web, News and Discover performance and has a separate Google News report. Discover is personalised and hard to attribute, so expect the weakest evidence there.
- Inspect the originals. Run URL Inspection, or the URL Inspection API for volume, on a sample of syndicated stories. If the Google-selected canonical is the partner’s URL, or the Page indexing report shows “Duplicate, Google chose different canonical than user”, you have a canonicalisation problem and a canonical or noindex request becomes urgent. If your original is indexed as its own canonical, you have a ranking competition problem.
- Check timing. Compare your publish time, the partner’s publish time and Googlebot’s first request for your URL in server logs. If the partner copy goes live before Google has fetched your original, fix discovery first (News sitemap entries, links from the homepage and section fronts). A delay clause buys little if Google has not fetched your story by the time it expires.
- Compare what ranks. Check whether the partner rewrites headlines, adds media or drops your attribution link. A rewritten headline can match queries your original misses.
- Size the loss. Compare clicks per story for syndicated and non-syndicated stories in the same sections over the same weeks. A partner outranking you on a handful of stories is not a pattern. The sized gap is your negotiating number, and it may show the problem sits with one partner or one section.
- Audit the partner’s pages. View source on a sample of partner copies for the robots meta tag, the canonical and the attribution link. Agreements and templates drift apart, so check rather than assume.
Terms worth writing into the agreement
- The mechanism in plain technical terms: a
robotsorgooglebotnoindex, or a canonical to the original URL, and which content types each applies to. - A delay window per content type: longer for exclusives and breaking news, shorter or none for evergreen pieces, where it protects little.
- A visible link to the original near the top of each copy.
- Whether the partner may rewrite headlines.
- Audit rights and a remedy if the terms are not implemented.
- A review date, because the partner’s reach and your search losses both shift over time.
A tiered deal can be easier to agree than a blanket one: noindex on exclusives and breaking news, canonical plus delay on the rest. Another option to table is a time-limited noindex that the partner lifts after a set number of days. No public evidence covers how that performs, and the partner’s copy competes for the long tail once indexed, so treat it as something to test.
How to test which concession is worth fighting for
You cannot randomise Google, but you can randomise stories. If a partner agrees to per-story terms, assign syndicated stories by a fixed rule (odd or even article ID, for example) to a delay or no-delay group for a few weeks. Comparing two partners with different terms is weaker, because their audiences and sites differ.
- Hypothesis: a delay window recovers most of the clicks your original loses to the partner’s copy.
- Primary metric: clicks to your original in the first 48 hours and first seven days, from Search Console page-level data for Web and News plus the Google News report.
- Secondary metrics: Discover clicks (read with caution), referral clicks from the partner’s attribution link, syndication revenue per story, and spot checks of Top Stories.
- Length: stories have short lives, so story count matters more than calendar time. A publisher syndicating dozens of stories a day collects hundreds of observations within a few weeks; one syndicating a handful a week needs months. Stratify by section and by breaking versus feature content.
- Confounders: major news events, algorithm updates and the partner’s homepage placement decisions, which you cannot see.
- Decision rule: if delayed stories close most of the gap to non-syndicated stories, delay is the cheaper concession. If the gap persists, push for noindex on the affected content types, or build the measured loss into the licence fee.
The position I would take
I would stop treating canonical requests as the fix. First confirm with URL Inspection whether your originals are indexed as their own canonicals, then measure the click gap per partner and section. Where the gap is large and you have leverage, ask for noindex on breaking and exclusive stories, with a Google-only noindex as the fallback. Where you lack leverage, negotiate a delay window and excerpt terms, measure the 48-hour gap on randomly assigned stories, and put the measured loss into the price of the deal.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.