Two different questions wearing the same word
Ask a distributor where your release went and you get a number: several hundred sites, or several thousand points of distribution. Ask in a year's time and the number is the same, because it was a count of an event, not of pages that still exist. The gap between those two things is the subject of this guide.
A pickup is an event in a distribution system. A feed was delivered to a partner, the partner's template rendered it, and something on the partner's server logged the fact. A findable page is a URL that exists, persists, is crawled, is stored, and can be returned to a person or an assistant that asks a question. Those two sets overlap on the day of publication and diverge sharply afterwards, and nobody in the chain has a commercial reason to tell you how fast.
The other guides on this site have taken the wire apart along other seams. Whether the links in a release pass ranking credit is settled: we checked 70 PR Newswire releases and every external link was nofollowed. What a wire will and will not accept is documented in its own guidelines, which is the subject of its own report. What the wires tell AI crawlers is in their robots files. Whether journalists read releases at all is a survey question, and whether an assistant will quote one is a retrieval question. None of those is the geography question. This is the geography question: where do the copies physically sit, and how long are they there?
Everything below comes from reads we performed on 26 September 2026, using a plain HTTP client against live servers, and from the public Common Crawl and Internet Archive indexes. Where a server refused us, we say so rather than reaching for a secondary account of what it would have said.
One release, traced as far as it could be traced
We started at the source of truth a distributor cannot hide: its own news sitemap. PR Newswire's sitemap-news.xml is a sitemap index of 18 pages. Page one held 99 releases with publication timestamps running from 16:05 Eastern on 25 September to 11:21 Eastern on 26 September — a shade over nineteen hours of output in one ninety-nine-item file. That is the raw material: a continuous, high-volume stream, indexed by the distributor for crawlers, turning over fast enough that a single page of the sitemap covers less than a day.
We then went looking for the copies. Because the distributors do not publish machine-readable partner lists, we worked the other way round: we took aggregators known to run wire feeds, fetched their own public listings, and matched slugs against the sitemap. Morningstar was the one that answered cleanly. It runs a dedicated section at morningstar.com/news/pr-newswire/, its listing page showed 50 items, and 35 of those 50 matched releases in the 99-item sitemap window we had just pulled. That is a live, verifiable syndication relationship, measured from both ends.
StockTitan answered halfway. Its news listing loaded and named Business Wire twenty times, PR Newswire seventeen times and ACCESS Newswire four times on a single page, which is about as clear a statement of a feed relationship as an aggregator can make. But every attempt to open an individual StockTitan article returned HTTP 403, so we could not inspect a copy there and will not describe one.
Everything else refused us. Business Wire returned an Akamai “Access Denied” page with HTTP 403 to a request for businesswire.com/robots.txt — not a release, not a paywalled asset, the file whose entire purpose is to be read by machines. GlobeNewswire failed the HTTP/2 connection with an INTERNAL_ERROR and then timed out over HTTP/1.1; we never read its robots.txt at all. ACCESS Newswire's own distribution and pricing pages returned 403. Associated Press's press-release section returned 403. Le Lézard, a long-standing PR Newswire partner, served a Cloudflare managed challenge. Business Insider's markets news listing loaded but named no wire service anywhere on the page we read.
That pattern is itself a finding, and it has a practical edge. If you are handed a placement report listing three hundred sites, you cannot check it. Neither can we. The infrastructure that counts the pickups is open to the distributor and closed to everyone else.
HTTP status returned to an ordinary request, 26 September 2026
Source: Media Morpheus Research live reads, 26 September 2026, using curl with a standard desktop browser user-agent string. Full endpoint list in the methodology note.
Bars encode outcome class, not a measured quantity: full width for a 200, reduced for a 403, minimal for a connection failure. A 403 here is a refusal to serve our client, not evidence the page does not exist.
What a partner copy says about itself
The one partner copy we could open properly is worth dwelling on, because almost every claim people make about syndicated releases is a claim about the HTML of pages like this one.
The release was a 25 September 2026 PR Newswire item about a cosmetic product launch. The wire's own page declared rel="canonical" pointing at itself, which is what you would expect. Morningstar's copy also declared rel="canonical" pointing at itself — at the Morningstar URL, not at prnewswire.com.
That single detail undoes a widespread assumption. People often say that syndicated wire copies canonicalise back to the wire and therefore cannot compete with it or dilute it. On this copy, they do not. Morningstar presents its rendering as an original page and leaves the duplicate question entirely to the search engine, which is exactly the clustering decision Google describes in its own documentation and which we set out in how long Google takes to index a new page. Google picks one representative of the cluster. Nothing in the partner copy's markup tells it which.
The link picture was almost empty, and in a way that matters. The string “nofollow” appeared zero times in Morningstar's HTML. Ten external anchors existed on the page; nine were Morningstar's own social and corporate destinations, and the tenth was a link back to the PR Newswire original. The release body, as rendered by Morningstar, contained no outbound links to the issuing company at all. So the nofollow question that dominates discussion of wire SEO does not even arise on this copy: there was nothing to qualify, because the partner template did not carry the body links through. On the wire's own page, the only nofollow in the document was on a telephone link. This is consistent with what our larger sample of release bodies found — most releases contain no external body links in the first place — and it means the link-equity framing is usually the wrong question to ask about syndication.
Two more differences between original and copy are worth recording. Morningstar's page yielded 1,202 words of extracted text; PR Newswire's yielded 1,592. And Morningstar wraps the release in an explicit disclaimer: third-party content “is offered for informational purposes only and is not endorsed, reviewed, or verified by Morningstar”, with no representation as to “accuracy, completeness, timeliness, or reliability”. If you are assembling coverage as evidence — for a visa petition, a funding round, a credibility page — that sentence is on the page, and a reader evaluating the citation will see it. The distinction between a wire copy and something an editor chose to run is the subject of wire release, contributed post or earned coverage, and a partner's own disclaimer makes the distinction for you.
Are the copies crawled? Yes — and that is not the problem
Indexability is cheap to establish. The Morningstar copy has no noindex, no robots meta restriction and a self-referential canonical: nothing about it forbids storage. So we went and looked at whether anyone had stored it, using Common Crawl's index — public, queryable by URL prefix, dated, and the corpus a great deal of downstream language-model and retrieval work draws on, which is the subject of its own guide.
They had. A single prefix query against the September 2026 index, CC-MAIN-2026-39, returned 492 records under morningstar.com/news/pr-newswire/, of which 267 answered HTTP 200. Partner copies are reachable, crawled and stored while they exist. Anyone claiming syndicated releases are invisible to machines is wrong, and we were wrong about it too before running the query.
A caution about the totals, because this instrument misled us once. Repeated queries with different pagination parameters returned wildly different record counts for the same prefix — 3,862, 16,007 and 30,534 for PR Newswire's own releases across three attempts with different parameters — so we do not treat any of these as a count of what the corpus contains. What is stable and reproducible is the status distribution within a single query's result set, because each record carries the HTTP status the crawler received. That is the measurement below, and rates are robust to sampling in a way that totals plainly are not.
The ninety-day cliff
Each Common Crawl index record includes the status code the crawler got when it fetched the URL. That turns the index into something better than our own probing: a log of what a well-behaved crawler found when it tried to read hundreds of partner copies, without us hammering anybody's server.
For PR Newswire's own release pages, the result is a flat line. Of 30,534 records returned in one query, 30,297 were HTTP 200 and 52 were HTTP 404 — a 0.2% loss rate, with the remainder redirects and a handful of 403s and 502s. StockTitan's news section was similar: of 2,296 records, 2,269 were 200, none were 404, and four were HTTP 410 Gone.
For Morningstar's copies of those same releases, the result is a cliff. Of 492 records, 222 — 45.1% — returned 404. And the 404s are not scattered. Sorting by the release date embedded in each URL, every datable copy dated on or before 10 June 2026 returned 404; the oldest copy that still answered 200 was dated 8 June 2026. Those fetches happened between 4 September and 16 September 2026, timestamps we read off the records themselves. That is a retention window of roughly ninety days, administered on a schedule, not gradual link rot.
Below the cliff, the older vintages are uniform: 2 of 2 copies from 2018, 2 of 2 from 2019, 2 of 2 from 2020, 3 of 3 from 2021, 7 of 7 from 2022, 4 of 4 from 2023, 27 of 27 from 2024 and 65 of 67 from 2025 returned 404. The two exceptions in 2025 were a 301 and a 302, which is to say redirects rather than surviving pages.
Our own live probing, done before we thought of using the crawler's logs, agrees: five 2015-dated Morningstar wire copies located through the Internet Archive's CDX index all returned 404 when requested on 26 September 2026, and the Archive's own last capture of one of them, on 22 July 2025, had already recorded a 404. We tried to extend that test across more vintages, at which point Morningstar began returning HTTP 202 with a zero-length body to our client, including for a 2026 URL we had successfully fetched twenty minutes earlier. That is rate limiting, it hit our control as hard as our test, and it is why the Common Crawl status data carries the argument instead.
HTTP 404 as a share of records, by the release date in the URL
Source: Common Crawl index API, collection CC-MAIN-2026-39, prefix query on morningstar.com/news/pr-newswire/, 492 records, queried by Media Morpheus Research on 26 September 2026. Fetch timestamps in those records run from 4 September to 16 September 2026.
Zero-value rows are drawn at minimum width so they remain visible. The 2025 figure is 65 of 67; the two non-404 records were redirects. Release dates are taken from the identifier in each URL, not from page content. This is one partner site and should not be generalised to others.
HTTP 404 as a share of records in one September 2026 Common Crawl query
Source: Common Crawl index API, collection CC-MAIN-2026-39, one prefix query per path, queried 26 September 2026. Statuses are the codes the crawler received; sample sizes are the records returned by each query.
Bars are scaled to the largest rate, so the lower two are deliberately close to invisible — that is the finding. StockTitan's clean result shows that partner-side deletion is a choice, not a property of aggregation.
StockTitan's zero is the important control here. It republishes the same wires and keeps its pages, which means the ninety-day cliff is Morningstar's editorial policy rather than an inevitable feature of syndication. You cannot predict from the outside which behaviour a given partner has, and no distributor's report distinguishes them.
How a pickup gets counted
While comparing the two copies of the same release we found something neither distributor documents at a URL we could open, and it explains the arithmetic of a pickup report.
Morningstar's copy makes a request to edge.prnewswire.com/c/img/favicon.png with two parameters: sn=CN51341 and sd=2026-09-25. The first is the release identifier — it matches the 20260925cn51341 segment of Morningstar's own URL. The second is the publication date. PR Newswire's own page for the same release does not make that request. Fetching the URL ourselves returned HTTP 302 with a zero-length body, redirecting to a static image path.
A request like that, embedded in the syndicated HTML and carrying the release ID, is a measurement instrument. Every time a partner's rendering of the release loads in a browser, the distributor's edge servers hear about it. That is a perfectly reasonable way to build a distribution report, and it is worth understanding precisely what such a report can and cannot contain. It can tell you a partner rendered your release and how often. It cannot tell you the page persisted, and on the evidence above it will not persist. We are reading a mechanism from its artefacts here rather than quoting documentation, because the documentation is not public; treat the inference as ours.
This is the same distinction at work in what an impression means, and it is why measuring whether a placement worked has to start by asking what the counted unit is. A pickup count is a delivery metric with a shelf life. The traffic question is separate and generally worse, as what happens to placement traffic sets out.
Why you cannot check the partner list
We set out to read the distributors' own published network documentation, as the honest way to answer a question about their networks. We largely failed, and the shape of the failure is informative.
PR Newswire is the most open of the four. Its robots.txt disallows only a handful of template and widget paths and then advertises a long list of sitemaps in fourteen languages. It also publishes an llms.txt, described in robots.txt under the heading “AI Context File”, which introduces the service as “the industry-leading global network for news distribution”, points at its latest-news, multimedia and public-company feeds, and closes by telling “large-scale crawlers” to “refer to our Sitemap Index” for the full corpus. Note what that file is doing: it steers machine readers to the distributor's copies. Nothing in it mentions a partner network. Whether anyone publishes such files at all is the subject of does anyone actually publish llms.txt, and what the wires tell AI crawlers specifically is in what wire services allow AI crawlers.
ACCESS Newswire's robots.txt is readable and modest: a short disallow list covering admin, API, list and RSS paths, and a Crawl-delay: 5. Its pricing and distribution pages, where a partner list would live, returned 403. Business Wire refused our client outright. GlobeNewswire's server would not complete a connection.
There is one distributor whose description of its own network is legally compelled to exist, because it is a public company. ACCESS Newswire Inc. (NYSE American: ACCS) filed its annual report for the year to 31 December 2025 on 19 March 2026. The relevant sentence is: “We continue to expand our distribution network, refine targeting capabilities, and enhance analytics reporting to maximize impact.” That is the whole of it. The filing names no partner site, gives no count of outlets, and the single appearance of the word “outlets” in 31,818 words describes something else entirely — its investor-relations content network, aggregated from “most of the major exchanges and news distribution outlets around the world”. The same filing reports total 2025 revenue of $22,619,000, down 2% from $23,057,000.
A distributor that must describe its business to the SEC under penalty of securities liability describes its distribution network qualitatively and nothing more. That is not an accusation; partner lists are commercially sensitive and change constantly. But it does mean that when a placement report tells you a release reached four hundred sites, the number is unauditable by construction, and should be read as a sales figure rather than a measurement. Why a filing is nonetheless a more reliable object than a press release is the subject of do investor materials get cited by AI, and the pricing claims around distribution are collected in what a press placement costs.
Methodology and limitations
All figures come from reads we performed on 26 September 2026 using curl with a standard desktop browser user-agent string, and from the Common Crawl and Internet Archive public indexes. The live endpoints were: prnewswire.com robots.txt, llms.txt, sitemap-news.xml (index and page 1) and one release page; accessnewswire.com robots.txt, pricing and press-release-distribution pages; businesswire.com robots.txt; globenewswire.com robots.txt and newsroom; morningstar.com robots.txt, the /news/pr-newswire/ listing, one current copy and ten archival copies; stocktitan.net robots.txt, news listing and one article; markets.businessinsider.com robots.txt and news listing; apnews.com press-release section; lelezard.com robots.txt; edge.prnewswire.com beacon URL; index.commoncrawl.org collinfo and the CC-MAIN-2026-39 index API; web.archive.org CDX API; sec.gov robots.txt, EDGAR company search and the ACCESS Newswire 10-K for fiscal 2025. The Common Crawl protocol was a single request per path with matchType=prefix, output=json, pageSize=1000 and page=0, counting the records returned and their status codes. Seven limitations should temper every number above. First, we inspected the markup of exactly one partner copy in full; canonical behaviour, link handling and text extraction differ per partner template, and one copy is an existence proof, not a distribution. Second, three of the four distributors and several partners refused our client, so the footprint we mapped is biased towards hosts that permit anonymous reads, and Business Wire's and GlobeNewswire's partner behaviour is genuinely unknown to us. Third, the Common Crawl record totals are not counts of the corpus: repeated queries with different pagination parameters returned 3,862, 16,007 and 30,534 records for the same PR Newswire prefix, which is why we report status rates within a single result set and say so explicitly. Fourth, the retention finding is one partner site. StockTitan's contrasting zero shows the behaviour varies, and we have no basis for estimating how common a ninety-day purge is across a distribution network we cannot enumerate. Fifth, release dates were parsed from identifiers in URLs rather than from page content, which is reliable for Morningstar's scheme and would not be for a partner using opaque slugs. Sixth, the pickup-beacon account is our inference from an observed HTTP request and a matching release identifier, not from documentation, which we could not open. Seventh, we did not buy a release; a purchased distribution with a paid visibility report might well show the partner list we could not read, and the argument above concerns what an outsider can independently verify, which is the standard that matters when a placement is being used as evidence.
Conclusion
A wire release ends up in three places and they have three different lifespans. It ends up on the distributor's own domain, where it is canonical, sitemapped and effectively permanent: a 0.2% loss rate across 30,534 crawler fetches. It ends up on partner sites as a feed rendering — shorter than the original, self-canonicalising, stripped of the issuer's body links, wrapped in a disclaimer that the host has verified none of it, crawled while it lives, and on the one partner we could measure, deleted after about ninety days, with every copy dated before 10 June 2026 already returning 404 to a crawler working in early September. And it ends up as a row in a report, counted by a beacon that fires when a partner's template renders, which records a delivery on a day rather than a page in a year. The practical consequences are short. Cite the wire's own URL, because it is the copy that survives. Treat a pickup list as a dated sales artefact, not an inventory of live pages, and check a sample of the URLs before putting the list in front of anyone who matters. Archive anything you intend to rely on, on the day it publishes. And note that StockTitan kept every page while Morningstar kept ninety days of them, which means partner persistence is a policy you cannot see from outside and nobody in the chain will tell you.
Frequently Asked Questions
Where does a press release actually get published?
On the distributor's own domain, durably; on partner sites, temporarily. Common Crawl's September 2026 crawl requested 30,534 URLs under prnewswire.com/news-releases/ and 30,297 of them answered with HTTP 200 — a 0.2% loss rate. The same crawl requested 492 URLs under morningstar.com/news/pr-newswire/, where Morningstar republishes PR Newswire releases in full, and 222 of them returned 404. Both copies are real. Only one of them lasts.
How long do partner copies of a release stay online?
On the one partner we could measure properly, about 90 days. Every datable Morningstar copy with a release date on or before 10 June 2026 returned 404 to a crawler fetching between 4 and 16 September 2026, and the oldest copy that still answered was dated 8 June 2026. Copies from 2018 through 2024 were gone without exception. That is a retention window, not a decay curve: the copies are not rotting, they are being deleted on a schedule.
Do partner copies of a press release carry a canonical tag pointing back to the wire?
The one we could read did not. Morningstar's copy of a 25 September 2026 PR Newswire release carried <link rel="canonical"> pointing at its own Morningstar URL, not at prnewswire.com. It does link to the PR Newswire original in the body, but as an ordinary link, not a canonical declaration. Each copy therefore presents itself to search engines as an original page, and the duplicate-clustering decision is left to the search engine.
Are syndicated copies of a release crawled and indexable?
Yes, while they exist. Common Crawl retrieved 267 live Morningstar copies in a single September 2026 prefix query, so partner copies are visibly reachable. The copy we inspected had no noindex, no robots restriction and no nofollow anywhere in its HTML. Invisibility is not the problem with a pickup. Impermanence is.
What is the difference between a pickup and a page a reader could find?
A pickup is a machine event on a partner's server. A findable page is a URL a human or an assistant can reach. On the Morningstar copy we read, the release HTML carried a request to edge.prnewswire.com/c/img/favicon.png with the release ID and publication date in the querystring, and that request was absent from PR Newswire's own page. A beacon like that fires whenever the partner's page renders, which is enough to count a pickup. It says nothing about whether the page will exist in three months.
Can I verify a distributor's partner list myself?
Largely not, on the evidence of a day spent trying. Business Wire returned HTTP 403 to a plain request for its robots.txt. GlobeNewswire's server failed the connection twice. ACCESS Newswire's distribution and pricing pages returned 403. The one distributor that files with the SEC describes its network in a 10-K only as something it continues to “expand”, naming no partner and giving no count. The lists exist inside sales materials, not at public URLs an outsider can read.
Which copy should I cite if the placement is evidence?
The wire's own URL, every time, and archive it anyway. The distributor's page had a 0.2% loss rate across 30,534 crawler fetches; the partner copy of the same release has a roughly 90-day life. A pickup list is a record of delivery on a particular day, and citing one in a document that will be read in a year means citing URLs that will mostly have gone.
Sources
- operator Common Crawl, index API, collection CC-MAIN-2026-39 (September 2026 Index). Prefix queries on morningstar.com/news/pr-newswire/ (492 records: 267 × 200, 222 × 404, 2 × 301, 1 × 302), prnewswire.com/news-releases/ (30,534 records: 30,297 × 200, 52 × 404, 107 × 301, 72 × 308, 4 × 403, 2 × 502) and stocktitan.net/news/ (2,296 records: 2,269 × 200, 18 × 301, 5 × 429, 4 × 410). Fetch timestamps 4–16 September 2026. Queried 26 September 2026.
- live read PR Newswire, robots.txt and llms.txt. The “AI Context File” heading; the instruction that “large-scale crawlers should refer to our Sitemap Index”. Read 26 September 2026.
- live read PR Newswire, sitemap-news.xml. Sitemap index of 18 pages; page 1 held 99 releases timestamped 25 September 16:05 ET to 26 September 11:21 ET. Read 26 September 2026.
- live read ACCESS Newswire, robots.txt. Disallow list and Crawl-delay: 5. Read 26 September 2026. Its press-release distribution page returned HTTP 403 to our client.
- live read Business Wire, robots.txt. Returned an Akamai “Access Denied” page, HTTP 403, on 26 September 2026. Cited only as a refusal; we did not read its contents.
- live read GlobeNewswire, robots.txt. HTTP/2 stream INTERNAL_ERROR, then a timeout over HTTP/1.1, on 26 September 2026. Not read.
- aggregator Morningstar, PR Newswire news section, and one copy, MAYPHARM Unveils HYALMASS AQUATOX, 25 September 2026. Self-referential canonical; no noindex; zero occurrences of “nofollow”; 1,202 words extracted; third-party content disclaimer; the edge.prnewswire.com beacon. Read 26 September 2026.
- live read PR Newswire, the same release on the wire's own site. Self-referential canonical; 1,592 words extracted; a single nofollow, on a telephone link; no beacon request. Read 26 September 2026.
- aggregator StockTitan, news listing. Named Business Wire 20 times, PR Newswire 17 and ACCESS Newswire 4 on one page. Individual article pages returned HTTP 403 to our client. Read 26 September 2026.
- operator Internet Archive, CDX server API. Enumerated Morningstar wire-copy URLs from 2015 onwards; the 22 July 2025 capture of one 2015 URL carries status code 404. Queried 26 September 2026.
- filing ACCESS Newswire Inc., Annual Report on Form 10-K for the year ended 31 December 2025, filed 19 March 2026 (CIK 843006, NYSE American: ACCS). “We continue to expand our distribution network…”; the single use of “outlets”; total revenue $22,619,000 against $23,057,000. Read 26 September 2026.
- documentation U.S. Securities and Exchange Commission, robots.txt. Allow: /Archives/edgar/data, which is what permitted the filing read above. Read 26 September 2026.