Two ways to ask the question
“Is a tier-one feature still worth it” is unanswerable as posed, because worth depends on the job. A feature can simultaneously be excellent for one purpose and useless for another, and the usual sales conversation collapses these into a single price.
We can make it answerable by picking measures that are public, checkable and independent of each other. We used two. The first is how Wikipedia's perennial sources list rates the outlet, which is the best available public proxy for whether a source can support claims about you in the most scrutinised reference work on the web.
The second is how many AI crawlers the outlet blocks in its robots.txt, which is a direct proxy for whether AI systems are permitted to read the coverage at all. One measure is about credibility; the other is about access.
Neither is a complete account of value. Neither captures human readership, referral traffic, sales credibility or the recruiting effect of a recognisable logo. They are two things that can be measured rather than asserted, which is more than most tier definitions offer.
How Wikipedia rates them
Thirteen of the fifteen outlets we checked are rated generally reliable: The New York Times, The Washington Post, The Wall Street Journal, Reuters, the Associated Press, The Guardian, CNN, Bloomberg, Axios, Politico, NPR, The Atlantic and Wired.
Two are not. Forbes sits at no consensus, and so does Business Insider, whose entry records that in a 2025 request for comment “editors determined that any LLM-generated writing disclosed as written using AI is generally unreliable”, with no consensus on the publication overall.
Several of the thirteen carry qualifications that matter more than the headline status. Bloomberg News is generally reliable, but Bloomberg company and executive profiles are rated separately at no consensus, “generally considered to be based on company press releases”, with consensus that they “should not be used to establish notability”. The Guardian's blogs are rated separately from its reporting.
So on the credibility measure, the tier-one label mostly holds — with the significant exception that the two outlets most commonly sold as accessible tier-one placements are precisely the two that fail it.
Generally reliable is also narrower than it sounds, because the status attaches to a source within its areas of expertise rather than universally. The New York Times entry notes that popular press sources “should generally not be used to support medical claims”. The Atlantic's records that it “does not always clearly delineate between reporting and opinion content”, with its Ideas column governed by the opinion guideline. Reuters carries the direct warning that “press releases published by Reuters are not automatically reliable” — syndication through a trusted wire does not launder the underlying document.
One rule in this set works strongly in a buyer's favour and is rarely mentioned. Both wire services carry explicit syndication provisions: “Syndicated reports from the Associated Press that are published in other sources are also considered generally reliable”, and Reuters carries the same. A wire story about you retains its standing wherever it is republished, which means the pickup matters less than the original placement — the opposite of how syndication is usually sold.
How many AI crawlers they block
We read each outlet's robots.txt on 19 September 2026 and classified six named AI user agents: GPTBot, OAI-SearchBot, ChatGPT-User, CCBot, ClaudeBot and Google-Extended. Politico returned HTTP 403 and is excluded.
The spread is extreme. Four outlets block all six. Two block none. The rest sit between, and the pattern within an outlet is often deliberate: several block crawlers associated with training corpora while permitting search-facing agents.
The Washington Post's carve-outs are the most revealing detail. It bars AI from journalism while admitting it to corporate and advertising paths, which tells you the policy is about commercial control of content rather than a blanket objection to machine reading.
Forbes presents the sharpest split of all: it blocks GPTBot while explicitly allowing OAI-SearchBot. The same coverage is withheld from one OpenAI agent and offered to another.
The Atlantic shows the cleanest version of the split. It allows GPTBot, OAI-SearchBot and ChatGPT-User while blocking CCBot, ClaudeBot and Google-Extended — permitting the agents associated with answering user queries and barring those associated with assembling training corpora. Axios inverts one piece of it by explicitly allowing Google-Extended while blocking CCBot. These are not blanket positions but negotiated ones, and they change.
Of six named AI user agents blocked in robots.txt, 19 September 2026
Source: Media Morpheus, September 2026. Original measurement of robots.txt across fifteen news domains; Politico returned HTTP 403 and is excluded.
Agents tested: GPTBot, OAI-SearchBot, ChatGPT-User, CCBot, ClaudeBot, Google-Extended. robots.txt records stated permission, not crawler behaviour.
The two measures disagree
Put side by side, the measures almost cancel out. The New York Times, CNN, Bloomberg and NPR are all generally reliable and all block every agent tested. The Wall Street Journal and Reuters are generally reliable and block none.
That means credibility and machine accessibility are close to independent properties. Knowing that an outlet is a trusted source tells you very little about whether an AI assistant is permitted to read your coverage there, and knowing that an outlet is open to crawlers tells you nothing about whether a citation from it will survive scrutiny.
Forbes illustrates the incoherence of a single tier label better than any other entry. It is partially open to AI crawlers, rated no consensus as a source, and split three ways by Wikipedia according to who wrote the article. One outlet, three different answers depending on the question.
The important caveat is that robots.txt records stated permission rather than observed behaviour. It establishes what a publisher has asked crawlers to do, not what any crawler actually did, and a blocked page can still be described in an AI answer from other sources that discuss it.
So what is a feature actually worth?
It depends which of three jobs you are buying, and they are genuinely different purchases. For human credibility — a logo a prospect recognises, a link in a pitch deck, a signal to a recruiter — tier-one features retain essentially all their traditional value, and neither measure here touches that.
For encyclopaedic standing, the byline type matters more than the masthead. Staff reporting in any of the thirteen generally reliable outlets does the job. A contributed article in the same publication generally does not.
For AI retrieval, a feature in an outlet that blocks every agent may be substantially less visible to those systems than equivalent coverage somewhere more open, which is a genuinely new consideration and one that inverts part of the traditional hierarchy.
The sensible reading is not that tier-one features have stopped mattering. It is that a single tier label now bundles three separable outcomes that no longer move together, and a buyer should specify which one they are paying for.
One consequence deserves stating directly, because it inverts a long-standing assumption. For the AI retrieval job specifically, some of the most prestigious destinations are among the least useful, since four of the outlets in our set bar every agent we tested. A buyer optimising for presence in AI answers and a buyer optimising for board-level recognition should, on this evidence, sometimes choose different publications.
What this means for anyone buying coverage
Ask the vendor which of the three jobs the placement is for, and be sceptical of any answer that claims all three at once. The evidence here shows the outlets that excel at one frequently do not excel at another.
Both measures are free to check before you commission anything. An outlet's perennial sources status is public, and its robots.txt is one request away. That is a more useful pre-purchase check than a media kit.
Be particularly careful with the two outlets that fail the credibility measure, because they are also the two most heavily marketed as attainable tier-one placements. Attainability and standing are related for a reason.
Media Morpheus sells media and podcast placement. Findings that complicate the value of tier-one features complicate our own pricing, and they are published here as measured rather than filtered for commercial convenience.
Methodology and limitations
Two datasets are combined. Source status comes from our parse of the raw wikitext of Wikipedia:Reliable sources/Perennial sources on 18 September 2026, reading the status of each outlet's entry. Crawler posture comes from our measurement of robots.txt across fifteen news domains on 19 September 2026, following redirects and parsing grouped user-agent blocks, since consecutive User-agent lines share the directive block that follows. Politico returned HTTP 403 from our vantage point and is excluded from crawler figures; that is a retrieval failure, not evidence of a block. Limitations: robots.txt states permission and not behaviour, and does not establish that any crawler obeyed or ignored it; an outlet blocking a crawler does not prevent it being described in AI answers via other sources. The perennial sources list is explicitly not a policy or guideline and records consensus at a point in time. Fifteen outlets is a purposive selection of widely recognised publications, not a random sample, and no causal claim is made from the pairing of the two measures.
Conclusion
Tier one is one label covering at least three separable outcomes. Measured against Wikipedia's source classifications, thirteen of fifteen major outlets hold up, and the two that do not are Forbes and Business Insider. Measured against AI crawler access, the same set ranges from four outlets blocking every agent tested to two blocking none. The measures barely correlate, so an outlet's credibility as a source is close to no guide to whether AI systems may read your coverage there. Features remain worth buying for the jobs they actually do. They are no longer worth buying on the assumption that one placement satisfies every objective at once.
Frequently Asked Questions
Which major outlets block the most AI crawlers?
Of fourteen readable domains on 19 September 2026, The New York Times, CNN, Bloomberg and NPR each blocked all six agents we tested. The Wall Street Journal and Reuters blocked none.
Are tier-one outlets automatically reliable sources for Wikipedia?
Mostly, but not universally. Thirteen of the fifteen we checked are generally reliable. Forbes and Business Insider are both rated no consensus, and some sections — such as Bloomberg's company profiles and The Guardian's blogs — are rated separately from the main publication.
Does blocking AI crawlers mean my coverage is invisible to AI?
Not necessarily. robots.txt records stated permission rather than crawler behaviour, and an assistant can still describe an event using other sources that discuss it. It does mean the outlet has asked those agents not to read the page directly.
Should I avoid outlets that block AI crawlers?
Only if AI retrieval is your objective. Those outlets are often the strongest for human credibility and encyclopaedic standing. The point is to decide which job the placement is doing rather than assuming one feature serves every purpose.
How can I check an outlet before commissioning coverage?
Both measures are public. Look up the outlet on Wikipedia's perennial sources list for its status, and request the domain's robots.txt to see which AI agents it names. Together they take a few minutes and tell you more than a media kit.
Sources
- vendor research Media Morpheus, September 2026. Original measurement of robots.txt across fifteen news domains on 19 September 2026, classifying six named AI user agents by grouped directive blocks. Method and limitations stated above.
- vendor research Media Morpheus, September 2026. Original parse of Wikipedia:Reliable sources/Perennial sources on 18 September 2026, reading the recorded status of each outlet's entry.
- official Wikipedia. Reliable sources/Perennial sources: the entries for the fifteen outlets, including the Business Insider 2025 request for comment and the Bloomberg profiles entry.
- official Google Search Central. Documentation of Google-Extended and the preview controls available to publishers.