Four things can happen, and only one of them is common
“What happens to my coverage when the journalist leaves?” is really four questions, and collapsing them is why the folklore on this subject is so pessimistic.
The article can be deleted, or survive. The byline can be stripped to a staff credit, or survive. The author page the byline links to can be deleted, redirected, orphaned or left exactly as it was. And the structured data naming the author can be removed, left stale, or updated to say the writer has gone. Those four things are governed by different systems inside a publisher, and they fail at very different rates.
The short version of what we found is that the first three are much more robust than people assume, the fourth is inconsistent in ways that matter for anyone relying on machine-readable authorship, and the catastrophic case is not a reporter resigning at all — it is the publication closing, which is a different guide's subject and a different order of magnitude of loss.
This matters because a byline is the load-bearing part of a media placement used as evidence. Immigration adjudicators, funding diligence processes and credibility pages all ask who wrote it. The site's broader argument for why an authored piece outperforms a mention is in author bylines, and what Google actually says about bylines as a credibility signal is in do author bylines affect credibility signals. Neither answers the decay question, so we went and measured it.
Do author pages survive? We tested 57
The design was deliberately blunt. We took Internet Archive captures of four technology publications' front pages from 14–15 September 2016, extracted every author-page link on them, and requested those URLs live on 26 September 2026. Front-page bylines in 2016 are, by construction, staff writers at that date; ten years is long enough that a good share of them have moved on. If author pages rotted, this is where it would show.
They did not rot. All 16 TechCrunch author pages returned HTTP 200, including pages belonging to an editor-in-chief who left in 2023 and several reporters who left years before that. Ars Technica returned 200 for 17 of 18; the single failure was arstechnica.com/author/andrew/, a first-name-only slug from an older naming scheme, which returned 404.
The Verge produced the most instructive result because it has changed its URL scheme since 2016, from /users/<username> to /authors/<real-name>. All eight old paths returned HTTP 308 permanent redirects. Six of them resolved to a live author page: /users/adirobertson/ to /authors/adi-robertson, /users/trixxy/ to /authors/thomas-ricker, and so on — including for staff who have since left. That is a publisher carrying identity across a platform migration, mapping obsolete usernames to real names, and keeping the old URLs working a decade later. Two of the eight ended at HTTP 410 Gone, which is worth noticing: 410 is the status a server returns to say a resource has been deliberately and permanently removed, as opposed to 404's “not found”. A publisher that bothers to return 410 is making a statement rather than leaving a hole.
VentureBeat refused us. All 15 of its author URLs returned HTTP 429 Too Many Requests on the first pass, and a slow retry at ten-second intervals returned 429 again. We have no data on VentureBeat and are not going to infer any.
Author URLs taken from 14–15 September 2016 front-page bylines, requested 26 September 2026
Source: Media Morpheus Research. Author URLs extracted from Internet Archive captures of techcrunch.com, arstechnica.com, theverge.com and venturebeat.com dated 14–15 September 2016; each URL requested live on 26 September 2026 with curl and a standard desktop user-agent string.
Bars are the within-publisher share of that publisher's sample, not a share of all 57 URLs. The VentureBeat row records a refusal, not a result. Four technology publications are not a sample of the news industry.
What happens to the byline on the article itself
The author page surviving is necessary but not sufficient. The other half of the question is whether the article still says who wrote it, and here the mechanism is on your side.
A byline is stored with the article. It is a field in the content record, written once at publication, not a live lookup against an employee directory. Nothing in a normal publishing system fires when an employment record changes. So the default is that nothing happens, and that is what we observed: a TechCrunch interview from 25 April 2021, written by an editor who has since left the publication, still shows his name in the visible byline, still hyperlinks that name to his author page, and still carries a Person node in the page's structured data naming him with a url pointing at the same author page.
One markup detail is worth flagging because it is a small bug with a large audience. On that article the Person node's identifier is minted under an image namespace — techcrunch.com/#/schema/person/image/… — and the article's author property references that identifier rather than a person-namespaced one. A consumer following the reference still arrives at a node with the correct name and URL, so the practical damage is nil, but it is a reminder that author markup on large sites is generated by templates nobody audits. Whether any of this markup affects how the coverage performs is a separate question, examined in does article markup help coverage rank.
Two cases do strip a byline, and both are editorial decisions rather than decay. The first is a correction or retraction that changes the attribution, which publishers handle under published policies — the territory of how publications handle corrections. The second is an unpublishing request, which is the subject of can a placement be removed. Neither is triggered by somebody changing jobs.
What the author schema says, and what it quietly stops saying
The interesting decay is not in the page or the byline. It is in the machine-readable claim about who the author is.
Google's guidance here is clear about the intent and silent about the endgame. Its article structured data documentation describes the author url property as “a link to a web page that uniquely identifies the author of the article — for example, the author's social media page, an 'about me' page, or a bio page”, notes that “you can use the sameAs property as an alternative” and that Google “can understand both sameAs and url when disambiguating authors”, and adds that “if the URL is an internal profile page, we recommend marking up that author using profile page structured data”. The ProfilePage documentation says the markup “helps Google Search understand the creators that post in an online community” and lists “an author page on a news site” as a valid use. Neither page says a word about what should happen to any of it when the author leaves. That is a genuine gap in the documentation rather than an oversight we can fill: there is no published guidance, and we are not going to invent a best practice and attribute it to Google.
What publishers do in the absence of guidance splits sharply.
The Verge updates the claim. Its author pages emit a Person object with a jobTitle, and the value is maintained: “Senior Editor, Tech & Policy” for a current staffer, “Deputy Editor” for another, and “Former Senior Reporter” for a reporter who left for another publication. The page still resolves, still carries her photograph and her archive, and now states in structured data that she is former. That is the behaviour to hope for: the historical record is preserved and the present-tense claim is corrected.
TechCrunch does not. A departed editor's page emits the same ProfilePage and Person markup as a current staffer's, with no employment status anywhere. We went looking for a proxy — the sameAs block that anchors a writer to their external profiles — on the theory that it might be pruned on departure. It is not a usable signal, because it is mostly absent either way. Across 19 TechCrunch author pages we parsed, only 4 carried any sameAs value at all, and the pages missing it included reporters who are demonstrably still on staff. Our hypothesis was wrong and the correction is worth stating plainly: on TechCrunch, the presence or absence of identity markup tells you nothing about whether the writer still works there.
The consequence is narrow but real. If your placement's value rests on the author being a verifiable, credentialed person — the framing behind how an entity enters the knowledge graph and the founder authority footprint — then the external links that make the author verifiable are exactly the part of the page most likely to be missing, and their absence is not something you can fix or predict.
Author pages parsed on 26 September 2026
Source: Media Morpheus Research, JSON-LD parsed from live author pages on techcrunch.com and theverge.com, 26 September 2026. Google's recommendations for the url and sameAs author properties are from Google Search Central's article structured data documentation, read the same day.
Bars are within-publisher proportions; the zero row is drawn at minimum width to stay visible. Three Verge pages is an illustration of a practice, not a measurement of how common it is.
When it is the masthead, not the journalist
Everything above concerns a reporter leaving a publication that continues. The case people conflate with it — a section being cut, a masthead being dissolved, a title being closed — behaves completely differently, and nothing about author-page durability carries over.
We checked what three closed publications now serve. Protocol, the technology title that closed in November 2022, returns HTTP 302 from its homepage to atom.com/name/Protocol, a domain-name marketplace listing, and returns 404 for an author path. Every byline ever published there now points, at best, at a page offering the domain for sale. The Messenger, which closed abruptly in early 2024, returns 200 for a 2,483-byte page consisting of its old tagline and an email address. Gawker returns 200 for a 16KB placeholder whose entire copy is “It's a website on the internet”, an email sign-up box and a 2026 copyright line. BuzzFeed News returned HTTP 406 to our client and we cannot describe it.
In none of those cases is there an author page to be orphaned or a byline to be stripped, because there is no archive. This is the same pattern the Pew Research Center measured across the whole web, and their two figures separate the two risks precisely: 16% of the pages in their sample were individually inaccessible on root domains that still worked, while 9% were gone because the entire root domain had failed. The first number is the ordinary churn of a functioning publisher. The second is what happened to Protocol. Pew was deliberately conservative, counting a page as lost only when it returned one of nine error codes that “definitively indicate that the page and/or its host server no longer exist or have become nonfunctional”, so these are floors rather than estimates. The wider decay picture, including the finding that heavily-trafficked news sites decay at the same rate as obscure ones, is in how long a placement stays findable.
Homepage response, requested 26 September 2026
Source: Media Morpheus Research live reads of gawker.com, themessenger.com, protocol.com and buzzfeednews.com, 26 September 2026. Byte counts are the transferred response bodies.
Bars encode the size of what is still served, scaled to the largest, which is a rough proxy for how much of a site remains rather than a measured quantity. A 200 status on a holding page is not an archive.
The check nobody runs: can a machine still see it?
There is a failure mode that looks like nothing at all from a browser. The author page resolves, the byline links, the markup parses — and no machine building a public index of the web can read any of it.
We tried to measure whether these author pages appear in the September 2026 Common Crawl index and got zeros across the board for techcrunch.com/author/, theverge.com/authors/ and arstechnica.com/author/. That is not evidence of orphaning, and the reason is in the publishers' robots files. TechCrunch's robots.txt disallows CCBot outright, alongside anthropic-ai, ClaudeBot, Applebot-Extended, ChatGPT-User, Bytespider, cohere-ai and others. The Verge's robots.txt gives CCBot Allow: /sp/ and Disallow: /. The corpus is empty of those pages because the publishers said so, which is a policy outcome rather than a decay outcome, and it makes the measurement we wanted impossible for these particular sites. What that corpus is used for is in what Common Crawl is; the breadth of publisher blocking is in do news sites block AI crawlers and which AI crawlers publishers name.
The practical point stands regardless of the mechanism. If your evidence is “an assistant can confirm that a named journalist at a named publication wrote about us”, the persistence of the author page on the publisher's server is not the binding constraint. Whether any machine is permitted to read it is. And that permission can be revoked in a single line of a robots file, at any time, with no notice and no trace on the page.
If the placement is evidence, act at publication
The operational advice falls out of the findings and it is short.
Capture three URLs, not one. The article, the author page as it stands that day, and the publication's masthead or staff listing. The article is the claim; the author page is the credential; the masthead is the proof that the publication existed with that staff at that date. Author pages are durable but their contents are not, and the biography and profile links are the part most likely to thin out.
Cite the author page by URL and by capture. Where the publisher states employment status in a machine-readable way, as The Verge does, a dated capture preserves the distinction between “Senior Reporter” at the time of writing and “Former Senior Reporter” now. Where the publisher says nothing, as TechCrunch does not, your capture is the only record of what the page asserted.
Do not treat a departure as a defect in the coverage. On the evidence above, an article written by a journalist who has since moved on is intact: same URL, same byline, same author link, same structured data. Reviewers who work with published material are used to this; the requirement in evidence frameworks is that the material carry a title, a date and an author, which it still does. That standard is set out in EB-1 media evidence and the criteria-level detail in EB-1A press criteria.
Worry about the publication, not the person. A reporter leaving costs you a biography. A publication closing costs you everything, and gives no notice. If a placement is load-bearing and the outlet is fragile, the archive is the whole insurance policy.
Methodology and limitations
The author-page test worked as follows. We fetched Internet Archive captures of techcrunch.com, arstechnica.com, theverge.com and venturebeat.com dated 14–15 September 2016 using the Wayback id_ raw-content endpoint, confirmed the vintage of each capture from date strings in the markup, extracted every author-page link on each front page, and requested those 57 URLs live on 26 September 2026 using curl with a standard desktop browser user-agent string, recording status codes and, where redirects occurred, the final effective URL. We separately parsed JSON-LD from 19 live TechCrunch author pages and 3 live Verge author pages, and from one 2021 TechCrunch article by a departed editor. We read Google Search Central's article structured data and ProfilePage documentation, Pew Research Center's May 2024 digital-decay study, the robots.txt of each publisher, and the current homepages of gawker.com, themessenger.com, protocol.com and buzzfeednews.com. Seven limitations matter. First, four technology publications on one platform generation are not a sample of the news industry; magazines, newspapers, wire services and trade titles run different systems and may behave nothing like this. Second, front-page bylines in 2016 skew towards senior staff, whose pages are the most likely to be maintained; a freelancer with three bylines is a different risk. Third, we have no data on VentureBeat at all, and none on BuzzFeed News, because both refused our client. Fourth, a 200 status is not a guarantee of content: we verified that author pages still list articles, but we did not systematically compare today's page contents against the 2016 versions, so “survives” means “resolves and still lists work”, not “is unchanged”. Fifth, the sameAs figures come from 19 pages on one publisher and should be read as a description of that publisher's markup, not an industry rate. Sixth, we could not test whether author pages are present in public web corpora, because the publishers block the crawler that builds the largest one. Seventh, we did not find any publisher documentation setting out what happens to an author page on departure; if such a policy exists in public we did not locate it, and we have described observed behaviour rather than stated policy throughout.
Conclusion
The decay people worry about is largely not happening, and the decay that is happening is somewhere else. Ten years after the fact, 16 of 16 TechCrunch author pages from September 2016 front-page bylines still return 200, 17 of 18 at Ars Technica do, and The Verge has carried six of eight obsolete username URLs across a platform migration onto live author pages while marking two as permanently gone with an honest 410. Bylines are article metadata and behave like it: a 2021 piece by a departed editor still names him, still links him and still says so in structured data. What thins out is the part that made the author verifiable rather than merely named — the biography, the external profile links, the sameAs block that only 4 of 19 TechCrunch pages carried in the first place — and the best practice we found anywhere is The Verge's, which keeps the page and changes the claim to “Former Senior Reporter”. Google's documentation asks for an author URL that “uniquely identifies the author” and says nothing at all about maintaining it. So the honest instruction is to stop planning for the journalist and start planning for the publication: a reporter moving on costs you a biography, while a masthead closing turns every byline you own into a 404 or, in Protocol's case, a redirect to a page where somebody is selling the domain.
Frequently Asked Questions
Does a journalist's author page get deleted when they leave?
Usually not. We took the staff bylines visible on four technology publications' front pages in September 2016 and requested those 57 author pages on 26 September 2026. All 16 TechCrunch pages returned HTTP 200, including several belonging to writers who left years ago. Seventeen of 18 Ars Technica pages returned 200. All eight of The Verge's old paths returned a permanent redirect, six of them landing on a live author page. On this evidence the author page is one of the more durable objects on a news site.
Does the byline come off the article?
Not on any publication we could check. A 2021 TechCrunch interview written by an editor who has since left still shows his name in the visible byline, still links that name to his author page, and still names him in the article's Person structured data. Bylines are stored as article metadata, not as a live employment lookup, so there is no mechanism that would strip one when somebody resigns.
What happens to the author schema markup?
It varies by publisher, and the best practice we found is explicit rather than silent. The Verge's author pages emit Person JSON-LD with a jobTitle, and for a departed reporter that value reads “Former Senior Reporter” against “Senior Editor, Tech & Policy” and “Deputy Editor” for current staff. TechCrunch emits ProfilePage and Person markup with no employment status at all, and the identity-anchoring sameAs property was present on only 4 of the 19 TechCrunch author pages we parsed.
Does a placement lose value as evidence when the reporter leaves?
Not for the reasons people fear. The article, byline and date survive, which is what an evidence reviewer needs. What can weaken is the author's own verifiable standing: the page that once carried a biography and links to the writer's other profiles may keep the name and drop everything else. If the point of your citation was that a named, credentialed journalist chose to write about you, keep a dated capture of the author page as it looked when it carried that biography.
What happens when the whole masthead is cut?
That is the real failure, and it is total rather than gradual. Protocol, closed in 2022, now redirects its homepage to a domain-name marketplace listing and returns 404 for author paths. The Messenger, closed in 2024, serves a 2.5KB holding page with an email address. Gawker serves a 16KB placeholder reading “It's a website on the internet.” In none of those cases is there an archive, an author page or a byline left to decay.
Can I tell from a publisher's site whether a writer still works there?
Sometimes, and only where the publisher chooses to say so. The Verge's structured data states it. TechCrunch's does not: a departed editor's page looks identical in structure to a current staffer's, and the absence of a sameAs block or a biography is no guide either, because current staff pages are missing those too. Absence of a signal is not evidence of departure.
What should I archive, and when?
Archive at publication, not when something goes wrong, and archive three things rather than one: the article URL, the author page as it stands that day, and the publication's own masthead or staff page. Pew's decay study counted a page as lost only when it returned one of nine error codes that definitively indicate the page or its host no longer exists, and still found 16% of pages individually inaccessible on domains that were otherwise working fine. The cheap moment to capture a page is while it still resolves.
Sources
- peer-reviewed Pew Research Center, May 2024. When Online Content Disappears. The 16% individually-inaccessible against 9% whole-domain split; the nine error codes that “definitively indicate” permanent loss; sample composition. Read 26 September 2026.
- documentation Google Search Central, Article (Article, NewsArticle, BlogPosting) structured data. The author url definition; sameAs as an alternative; the recommendation to use ProfilePage markup for internal profiles. Read 26 September 2026.
- documentation Google Search Central, Profile page (ProfilePage) structured data. Purpose; the single required mainEntity property; “an author page on a news site” as a listed use case; no guidance on removal. Read 26 September 2026.
- live read Internet Archive, Wayback Machine raw-content captures of techcrunch.com, theverge.com, arstechnica.com and venturebeat.com, 14–15 September 2016. Source of the 57 author URLs. Retrieved 26 September 2026.
- publisher TechCrunch author pages, including a departed editor's page and a current staffer's page. ProfilePage and Person JSON-LD; sameAs present on 4 of 19 pages parsed; no employment status. Read 26 September 2026.
- publisher TechCrunch, Interview: Apple executives on the 2021 iPad Pro, 25 April 2021. Visible byline, byline link and Person structured data intact for an author who has since left. Read 26 September 2026.
- publisher The Verge author pages: Ashley Carman (jobTitle “Former Senior Reporter”), Adi Robertson (“Senior Editor, Tech & Policy”) and Thomas Ricker (“Deputy Editor”). Read 26 September 2026.
- live read Publisher robots files: techcrunch.com/robots.txt, which disallows CCBot, anthropic-ai, ClaudeBot, Applebot-Extended, ChatGPT-User, Bytespider and cohere-ai; and theverge.com/robots.txt, which gives CCBot Allow: /sp/ and Disallow: /. Read 26 September 2026.
- operator Common Crawl, index API, collection CC-MAIN-2026-39. Zero records for techcrunch.com/author/, theverge.com/authors/ and arstechnica.com/author/, explained by the robots rules above. Queried 26 September 2026.
- publisher Closed publications, homepages requested 26 September 2026: protocol.com (HTTP 302 to a domain-name marketplace listing; 404 for an author path), themessenger.com (2,483-byte holding page), gawker.com (16KB placeholder) and buzzfeednews.com (HTTP 406, not read).