AI Has Made Volume a Poor Measure of Content’s Value
Generative AI lets companies produce far more content at lower cost, but more output doesn't necessarily bring more attention, engagement, or response.
Topics
News
- Indian Embassy in Jakarta Launches AI Avatar Dewi
- Anthropic Tests How Small AI Models Can Replicate Across IoT Devices
- Scientists Call for Global Oversight of Cyber-Capable AI
- Karnataka Eyes Green AI Model Amid Rising Power Needs
- Infosys Sees AI Driving Half Its Work Within Three Years
- FBI Arrests OpenAI Engineer Over Recording Device Planted in Office
Key Takeaways
01
AI has made it much cheaper and faster to produce polished content, so simply publishing more no longer gives a company much of an advantage.
02
Whether content performs depends on more than who or what wrote it. Relevance, expertise, distribution and the consequences of getting it wrong matter more.
03
Companies should use AI heavily for routine, low-risk work, keep people closely involved in higher-stakes content, and judge success by results rather than volume.
Pangram, which sells AI-detection software, found AI-generated text in 34% of the 178,672 English-language press releases it studied. The releases were distributed on PR Newswire between August 2025 and July 2026. The weekly share rose from 22% in August 2025 to a peak of 48% in June 2026.
AI has sharply lowered the cost of producing competent content, cutting the time and money needed to draft and translate it. But producing more does not show whether any of it performs.
India plans to widen access further. The government intends to train 15,000 creators and media professionals in AI through partnerships with Google and YouTube, the Reuters Institute noted in June. As polished content becomes easier to produce, companies can no longer assume that producing more of it will bring more attention, engagement or response.
Cheap Polish Has Stopped Signaling Effort
A well-prepared pitch once suggested that someone had spent time researching the publication and shaping an idea for it. AI has made that kind of polished outreach much easier to produce, and editors are now receiving far more pitches that sound alike.
Sonakshi Yajurvedi, a director at Jajabor Brand Consultancy, a Delhi-based public relations and brand-strategy firm, sees that in practice. “The frustration is real,” she told MIT SMR India. “Editors are drowning in pitches that sound alike because they were all written by the same tool with slightly different inputs.”
A well-prepared pitch once suggested that someone had spent time researching the publication and shaping an idea for it. AI has made that kind of polished outreach much easier to produce, and editors are now receiving far more pitches that sound alike.
Sonakshi Yajurvedi, a director at Jajabor Brand Consultancy, a Delhi-based public relations and brand-strategy firm, sees that in practice. “The frustration is real,” she told MIT SMR India. “Editors are drowning in pitches that sound alike because they were all written by the same tool with slightly different inputs.”
Featured, which runs three services connecting journalists with expert sources, including Help a Reporter Out, or HARO, received 1,394,740 pitches in response to 42,392 journalist requests seeking expert comments and sources for stories between June 2025 and July 2026. Pangram’s detector classified 292,895 of those pitches, about 21%, as fully AI-generated.
On Featured’s own platform, journalists selected about 10% of fully AI-generated responses, compared with about 33% of human-written responses.
On HARO, 35% of a pool of more than 2,500 journalists chose to switch on a filter that blocks fully AI-generated pitches. The filter is off by default. Pangram sells the detector and published the analysis using Featured’s data, so the findings carry a commercial interest.
Some editors are changing how they accept pitches. Hanna Raskin, founder of The Food Section, which covers food in the American South, moved freelance pitching to a phone line in September after receiving a flood of AI-generated and AI-assisted ideas.
Other companies have pushed automation further. In July, Futurism reported that UK public relations firm Movchan Agency had used at least 15 fabricated publicist identities, some with AI-generated headshots, to pitch journalists. Its clients said they had not known about the practice. The agency’s founder said real employees wrote the pitches and described the use of fake identities as poor judgment.
The examples point to the same practical problem. AI can make a pitch sound polished, but it cannot replace knowledge of the publication, the editor or the subject.
Yajurvedi puts it plainly. “The line between useful and noise is actually very simple: did a human think before hitting send?” she said. “AI can clean up a pitch, but it cannot know the context. That lives in a relationship.”
“Editors are drowning in pitches that sound alike because they were all written by the same tool with slightly different inputs.”
— Sonakshi Yajurvedi, Jajabor Brand Consultancy
AI Authorship Alone Does Not Decide What Gets Found
AI-assisted writing was already widespread across the web by 2025. Ahrefs, which sells search-marketing software, examined 900,000 English-language pages newly detected by its crawler in April that year, taking one page from each of 900,000 different websites. It found some AI-generated text on 74.2% of those pages. Only 2.5% were classified as entirely AI-generated.
That does not mean AI-written pages perform better in search. Studies looking at Google rankings have reached different conclusions.
Graphite, another search-marketing firm, examined Google results for 31,493 search terms collected in June 2025. It found that 86% of the articles appearing in those results were human-written and 14% were AI-generated. AI-written articles also tended to rank lower.
Ahrefs reached a different result in a separate study. It took 100,000 random search terms, collected the top 20 Google results for each and analyzed about 600,000 pages from that pool. It found almost no relationship between the amount of AI-generated text on a page and where that page ranked in Google.
The studies used different samples and methods, so they should not be read as evidence that Google either rewards or penalizes AI-written content. The clearer conclusion is that authorship alone does not determine whether a page ranks.
AI assistants create another route through which content gets found. Muck Rack, which sells public relations software, analyzed more than 25 million links cited by ChatGPT, Claude and Gemini for a May 2026 report. About 84% of those links came from earned sources, including news, research, government sites and other material that companies did not directly control. Paid and advertorial material accounted for just 0.3%. Journalism made up about 27% of all cited links, while 46% of responses to questions about industry trends cited at least one journalistic source.
Muck Rack also sells tools that track how brands appear in AI answers, so it has a commercial interest in the findings. Even with that qualification, the data makes one point clear. Publishing more pages on a company website does not guarantee that search engines or AI assistants will choose them when answering a query.
Another risk arises when many organizations use the same tools. Vagdevi Puranik, assistant professor in the postgraduate Department of Mass Communication and Journalism at St. Philomena’s College in Mysuru, sees it in newsrooms. “If too many newsrooms rely on the same systems, journalism risks developing a standardized voice, monotonous language, and repetitive framing,” she said. “Good reporting depends on critical thinking, dissent, and human sensitivity, qualities that get flattened with overuse of AI.”
The same concern applies to corporate content. Yajurvedi puts it more simply. “The result reads cleanly and says nothing that stays with you.”
The Cost of an Error Should Set the Level of Oversight
Not all content needs the same level of human review.
AI can handle much of the routine work companies produce at scale, including product descriptions, summaries, transcripts, standard translations and first drafts. These tasks still need checks, but errors are usually easier to spot and correct.
The stakes are different when content can affect investors, regulators, customers or a company’s reputation. Investor communications, technical claims, crisis statements, executive commentary and media responses need closer review because mistakes can have financial, legal or reputational consequences.
For Sonakshi Yajurvedi of Jajabor Brand Consultancy, the issue is who takes responsibility for the final decision.
“AI as editor-in-chief is a category error,” she said.
Some editorial decisions require someone to back a difficult call and stand by it if it proves unpopular. “These calls require integrity,” Yajurvedi said. “AI has no integrity. It has compliance.”
She makes the same point more directly when talking about accountability. “AI has no skin in the game. It has no reputation to protect, no source to lose, no community to answer to.”
Indian companies also have considerable room to decide how much human oversight AI-written material should receive. Amendments to the Information Technology Rules, in force since February 20, 2026, require platforms to label certain synthetically generated information. The definition covers audio, visual and audio-visual material, but not written text.
An AI-written press release, article or executive post therefore falls outside that specific labeling requirement. Other laws and regulations, including those governing misleading advertising and securities disclosures, still apply to what companies publish.
That leaves companies to decide when AI-written text needs disclosure, who must review it and which types of communication require named human sign-off.
“AI has no skin in the game. It has no reputation to protect, no source to lose, no community to answer to.”
—Sonakshi Yajurvedi, Jajabor Brand Consultancy
Implications by Role
C-Suite
Do not judge a content team by how much it produces. Press releases, posts and web pages show activity. They do not show whether the content reached the right audience or changed anything. When budgets are set for the fiscal year beginning April 2027, communications and marketing leaders should report outcomes alongside output. Those measures can include pitch response rates, citations of company data or experts, search and AI visibility, qualified engagement and conversions. Savings from AI can then be redirected toward original research, subject expertise and more time with journalists and customers.
Functional Leaders
Over the next quarter, separate routine content from material where mistakes carry greater consequences. AI can handle much of the high-volume work, such as summaries, translations and first drafts. Content involving factual claims, investors, regulators or the company’s public position should have a named human reviewer. Teams should also compare AI-assisted and human-led work using the same performance measures rather than assuming one will always work better.
Boards and Governance
India’s current labeling rules do not cover written AI-generated content, so companies need their own standards. At the next quarterly meeting, boards should ask management for a policy that states where AI can be used freely, where human approval is required and when substantial AI involvement should be disclosed. The policy should be reviewed as the technology and regulation change.
The evidence does not show that audiences prefer human-written content simply because a person wrote it. The search studies reach different conclusions, and several of the datasets come from companies with a commercial interest in the findings.
The evidence does show that producing more content does not guarantee better results. About a third of the press releases studied by Pangram contained AI-generated text. On Featured, fully AI-generated pitches were selected less often than human-written ones. AI assistants also drew most of their citations from sources outside a company’s own channels.
For managers, the question is therefore not how much content AI can produce. It is where AI saves time and where expertise, context and closer review still make a difference.
Research Highlight
This article draws on interviews with Sonakshi Yajurvedi of Jajabor Brand Consultancy and Vagdevi H.S. of St. Philomena’s College. It is supported by Pangram and Featured data on press releases and media pitches, Ahrefs and Graphite research on AI-generated web content and search rankings, Muck Rack’s analysis of more than 25 million AI citations, the Reuters Institute’s Digital News Report 2026, and India’s 2026 amendments to the Information Technology Rules.
