Syndicated content for accountants: what 370 matching articles tell us
I examined 1,915 pages published by accountancy firms to see how widely content was shared, how much it changed and which versions Google selected in a small search test. Syndicated content can be useful. But firms should be clear about what they expect it to achieve.
TL;DR
- Across seven selected firms classified article by article, 370 of 962 articles matched material published elsewhere. That’s 38.5%, but it isn’t a profession-wide estimate.
- Shared content has a perfectly sensible use: keeping clients informed without asking specialists to write every routine update.
- Its weakness is differentiation. If several organisations publish much the same article, your version may say little about your firm and search engines may select another copy.
- The Google pilot illustrated that concern, but the AI-answer tests didn’t establish a disadvantage for shared content.
- Grammarly’s AI detector produced inconsistent results across shared articles, comparison articles and known AI-assisted texts. It can’t establish authorship.
- If you want content that is specific to your firm, the important choice isn’t simply human versus AI. It’s whether the piece is commissioned for you, informed by your specialists and checked before publication.
Six findings in numbers
1,915 editorial pages published by 24 firms during the study year. The 25th had no blog and no published stories when checked.
370 of 962 articles matched material published elsewhere across seven selected firms: 291 exact shared-body matches and 79 with significant reviewed overlap.
28 of 31 sets of matching articles were recorded as having identical or substantially identical substantive copy. The other three still shared substantial passages but had been more extensively rewritten.
77+ further organisations carried related versions. Fifty-three were found by following matching articles originally found among the seven firms; 24 came from separate checks of the content-provider market and another cluster of matching articles. Not all of them bought syndicated content: the group includes business-network sharing, regional reuse and attributed press placements.
0 versus 4: none of the four shared versions I tested came first when I searched for the complete headline. All four comparison articles did; these were articles for which I hadn’t found matching copy elsewhere. It was one small test on one day.
76% versus zero: the highest AI-detector score in the whole test belonged to a comparison article, not a shared one. Six of the 14 shared articles scored zero. Detectors can’t establish who wrote an article or how it was produced.
The research
If I were discussing these results with a partner or marketing director, I’d start here: syndicated content isn’t inherently bad.
On the contrary, it can solve a real problem. Clients need clear updates on tax changes and reporting deadlines, and a small marketing team can’t ask an in-house specialist to write every one from scratch.
The difficulty comes when the same material is also expected to distinguish your firm from competitors, attract search traffic, demonstrate specialist knowledge and generate enquiries. A regular supply of articles doesn’t automatically do all four.
Shared material was a substantial part of the selected blogs
I began with a fixed cohort: the websites of 25 firms ranked from 76 to 100 in the Accountancy Age Top 50+50 Firms 2025. To avoid bias, I selected them before examining their websites for shared copy.
My initial finding was that, between 1 August 2025 and 31 July 2026, 24 of the websites published 1,915 editorial pages between them. One site had no blog or published news when I checked.
I searched for matching material elsewhere on the web and confirmed shared content at seven firms. I then classified all 962 articles they had published. Of these, 291 had exact shared-body matches and another 79 had significant overlap confirmed through review: 370 articles in total, or 38.5%.
Because I chose the seven websites for full classification only after I found shared material, 38.5% isn’t an estimate for either all 25 firms or for the profession. What that figure supports is a narrower conclusion – on some of the firms’ websites that I examined in detail, shared copy formed a substantial part of their published output.
In 28 of 31 comparisons, the main article was identical or substantially identical
The 31 weren’t another sample or 31 additional articles. Among the 370 were 31 sets of matching articles, each appearing on two, three or four of the seven firms’ websites. The sets involved four of the seven firms. I grouped the different versions and had the main text in each set reviewed separately.
In 21 comparisons, the main article text was identical. It was substantially identical in seven more. The remaining three retained a significant shared core. In the 21 identical comparisons, the recorded differences were limited to headings, formatting, dates and calls to action.
These matches also gave me starting points for searching beyond the original firms. By following their titles, distinctive phrases and matching errors, I found 53 further organisations carrying related versions.
Separate checks of the content-provider market and another cluster of matching articles identified 24 more. Together, the wider research found at least 77 further organisations beyond the seven fully classified firms.
That doesn’t mean all 77 bought syndicated content, or that every article travelled in the same direction. The wider evidence included content services, sharing within business networks, regional reuse, attributed press placements and cases where the source couldn’t be established. Matching language shows that pages are related. It doesn’t, by itself, prove who wrote the material, who supplied it or who published it first.
Shared copy has strengths. Distinctiveness isn’t one of them.
If the job is to explain a general change quickly, shared content may be a sound choice. The underlying rules don’t become more useful because 20 accountancy firms explain them in 20 different ways.
But once a client knows what changed, the questions become specific. What does this mean for our business? What should we do now? Those answers require your firm’s experience and judgement and – in particular – knowledge of its client base. A general article produced without your firm’s input can’t know which clients you want, cite cases your specialists have handled or tell readers what they’d advise.
When several firms publish the same article, Google may show somebody else’s copy
I didn’t find evidence that Google was penalising whole websites for publishing duplicated articles. The research wasn’t designed to test for that.
The practical issue is simpler. If several websites carry substantially the same article, Google has several versions to choose from. It can show somebody else’s without penalising your page or your website.
To explore this, I selected four shared articles. For each one, I chose another article from the same firm’s website where I hadn’t found matching copy elsewhere. That doesn’t prove those four comparison articles were original or exclusive. It simply gave me two sets to test.
First, I searched for each article’s complete headline. All four comparison articles appeared first. None of the four shared versions I was testing did. Google showed another version of the shared article instead. In one case, this was the page that the tested website had itself marked for search engines as the main version.
I then searched using broader phrases of the kind a prospective client might use. None of the four shared versions I was testing appeared directly. Another website’s version appeared in three of the four searches. Two of the four comparison articles appeared.
This was a small, signed-out test conducted on 28 August 2026. It covered four pairs of articles on different subjects. It can’t prove why Google chose those pages, where they’ll rank in future or whether any firm lost an enquiry.
Searching for a complete headline isn’t normal client behaviour. It was a way to see which version Google chose when asked to find a particular article.
The AI tests found no clear advantage or disadvantage for shared content
I gave four AI systems eight general accountancy questions. The questions didn’t name a firm or point the systems towards a particular website. Each system answered all eight, giving me 32 answers, which I saved.
Not one answer mentioned any of the firms whose articles I was testing. Nor did any of the other websites carrying the same shared articles appear. The comparison articles – where I hadn’t found matching copy elsewhere – were absent too.
I then ran a separate test asking AI systems to recommend accountancy firms. Across 16 runs, they made 79 recommendations. None went to the four firms included in this part of the research. One other firm carrying a confirmed shared article was recommended once.
These were small tests conducted at a particular point in time. They didn’t show that shared content performed worse than the comparison articles. But they didn’t show an advantage either. Neither set achieved useful visibility.
That inconclusive result matters. The evidence doesn’t support claiming that syndicated articles either help or harm a firm’s visibility in AI answers.
AI detectors gave mixed results – and no proof of authorship
I ran 24 texts through Grammarly’s AI detector while signed in to my account. These texts included 14 shared articles; six comparison articles for which I hadn’t found matching copy elsewhere; two substantial rewrites; plus two texts that I knew had been produced with ChatGPT assistance.
Grammarly gave eight of the 14 shared articles a non-zero “possible AI” score. The other six scored zero. Three of the six comparison articles also scored above zero. The highest result in the whole test – 76% – belonged to a comparison article, not a shared one. Both rewrites and both ChatGPT-assisted texts scored above zero.
None of the scores can establish who wrote an article, how it was produced or whether any firm used generative AI.
They do raise a more useful commissioning question. If a firm wants an article that won’t also appear unchanged on other websites, should it commission one specifically for its own audience?
That article might be written by a human, created with AI assistance or produced through a mixture of the two. Either approach can still result in generic or derivative copy, so the finished article needs to be checked for accuracy, similarity and proper sourcing.
The advantage is control. The firm can set the brief, provide its own examples, involve the right specialist and decide what the article needs to achieve. That gives the firm a better chance of publishing something another practice couldn’t use unchanged.
I haven’t named the firms or the content providers
That’s deliberate. Syndicated content can be useful, and this isn’t an exercise in embarrassing firms or content providers.
I’ve described the cohort precisely because the method should be checkable. That does mean a determined reader could work out which firms are in it. I’d ask them not to. Naming would change what this article is for, and none of the findings depend on who the firms are.
I’m interested in a simpler question: what is the content meant to achieve, and what evidence shows that it does? Search and AI answers are changing quickly, so familiar assumptions are worth testing.
The point is to help firms make better decisions, not to catch anyone out.
Only your results can show whether the content is worth paying for
From outside a firm, I can see what it published, where else the same material appeared and which version Google showed in a recorded search.
I can’t see how many clients read it, what they clicked in the newsletter, whether it produced enquiries or instructions, how much revenue followed or what the service cost.
That evidence remains inside a business: Google Search Console, website analytics, email reports, CRM records and call tracking. It also needs to be judged against whatever the content was supposed to achieve.
If the aim was to keep clients informed, and the email results show that clients read and value the updates, the service may be doing exactly what the firm bought it for. If the aim was distinctive search visibility or new business, the number of articles published won’t tell you whether either happened.
Start with what you want your content to achieve
Firms should decide what they want the content to achieve before signing up to any service. I’d ask four questions:
- What is its main job? Is it there to keep clients informed, maintain a regular publishing schedule, attract search traffic, demonstrate specialist knowledge or generate enquiries?
- What will make it ours? Which examples, opinions and recommendations will our specialists contribute?
- What exactly are we buying? How many other websites may receive the same article? What are we allowed to change? Which page will search engines be told to treat as the main version?
- How will we know whether it worked? Which figures will appear in the next report, and what result would justify renewing?
You may conclude that shared content is good value. You may decide to combine it with individually commissioned articles. Or you may discover that one content stream has been expected to do several jobs that nobody has measured.
Any of those conclusions is more useful than declaring syndicated content either good or bad.
If you’d like a second opinion, talk to me
If you’re reviewing a content arrangement, comparing agency proposals or trying to decide what your blog is actually for, invite me in for a chat.
I’m not going to tell you that syndicated content is bad. Instead, I can help you work out where it’s useful, where something written specifically for your firm would do a better job and whether it’s worth taking a different approach.
Methodology: I selected the cohort, set the research questions and definitions, checked points manually and directed the work. ChatGPT and Codex assisted with website collection, text matching, classification, testing, structured analysis and reconciliation. Claude was used for defined review and analysis tasks. AI-written summaries were treated as secondary analysis, not as primary evidence. Research materials included preserved webpages, screenshots, search and AI outputs, spreadsheets, decision records and evidence manifests. SHA-256 checksums were used to monitor file integrity; a later chain-of-custody check verified 954 recorded items without errors. Every figure in this article rests on that row-level evidence rather than on any AI-written summary.
A note on who wrote this. I used ChatGPT to write this article and ensured it stuck to the documented findings. I then edited it. On some occasions, only human-written copy will do. This isn’t one of them. I’m a copywriter, not Flaubert.
Know someone who needs to know more about syndicated content?
Share this article with them
