A new post on the British Council’s “Future of English” blog titled “Beyond the Score” conveys a now very familiar message.  Stuff like:

“…it’s important to recognise that the focus and robustness of assessment can vary significantly between different tests. Not all tests may evaluate the specific skills and language competencies needed for academic success in the same rigorous ways, while established tests such as IELTS (which the British Council helps deliver) continue to set the benchmark for trust and transparency.”

And:

“Furthermore, the concept of washback is crucial. This refers to the influence of the test on teaching and learning. If tests focus on a narrow range of skills or utilise formats that don’t reflect academic tasks (such as those often seen in shorter, computer-marked tests), this can lead to a narrowing of the curriculum and teaching practices, potentially disadvantaging students by not adequately preparing them for the full spectrum of academic language use.”

Like all of the other posts on this topic, the blog refers to the results of a single as-yet-unpublished survey of score users in the UK.

I share this mostly to emphasize how hard the IELTS partners are leaning into this message right now. There is an absolute ton of this sort of stuff coming out of Cambridge, the BC and IELTS HQ right now. That’s worth noting.

One might assume that this missive (and others before it) is a response to the growing influence of the Duolingo English Test. But it doesn’t seem like the IELTS partners are quite ready to specifically single out that test for criticism.

I really admire the folks at Cambridge and the British Council for having convictions and sticking by them. They truly believe that so-called “AI-First” tests are potentially harmful for learners and institutions. This isn’t just a marketing thing.

That said, this campaign isn’t going to work. The idea that persuasion and gravitas and one unpublished study can stand as a bulwark against AI-First testing is unconvincing.  We might ask all of the people who used to work at ETS how that approach fared in the USA.

Score users in the two biggest receiving markets (USA and Canada) are all-in on the Duolingo Test. The whole Ivy League accepts DET scores now.  All 15 of the biggest Canadian universities accept it as well. The DET team will keep hammering away at institutions in the UK (and marketing the test to UK-bound students) until they win in that country.

If Cambridge and the BC genuinely believe that the DET and similar tests are detrimental to schools and students, the onus is on them to create an alternative test which is equally attractive to students and score users, but lacks the perceived weaknesses of the DET test. If the problems they’ve described time and again truly exist, this is the only way to prevent them from becoming an even bigger concern.

And, needless to say, the current iteration of the IELTS is not such an alternative.

It appears that the paper version of the IELTS is now done in ink, all over the world.  I wrote a little while ago about this requirement being introduced in India.  Apparently it is now a global requirement.  Here’s an image from Facebook.

The Higher Education Policy Institute has published an article written by a managing director at Cambridge University Press & Assessment about what is described as “the shift to remote language testing that removes substantial human supervision from the process.”

Notes the author:

“While some may be excited by the prospect of an “AI-first” model of testing, we should pursue the best of both worlds – human oversight prioritised and empowered by AI. This means, for instance, human-proctored tests delivered in test centres that use tried and proven tech tools.”

And:

“Cambridge has been using and experimenting with AI for decades. We know in some circumstances that AI can be transformative in improving users’ experience. For the highest stakes assessments, innovation alone is no alternative to real human teaching, learning and understanding. And the higher the stakes, the more important human oversight becomes.”

Cambridge has been pushing back against newer tests with a bit more forcefulness in recent months (see also the “The impact of English language test choices for UK HE” report).

To my eye, the debate between at-home testing vs on-site testing is over, with supporters of at-home scoring a decisive victory. Indeed, Cambridge’s own at-home IELTS is widely accepted at schools across key receiving markets. But more importantly, it seems that most test takers really like the idea of at-home testing. Many of those who forgo it in favor of an on-site test do so out of fears that the maker of their chosen test stinks at delivering a seamless at-home product – not because of some love of the test center experience. As test makers get better at doing at-home testing, more test takers will pile into that option.

There might still be room for a robust debate about the merits of synchronous online proctoring (that is, a proctor watches as you take the test) vs asynchronous online proctoring (a proctor watches a video of your test after the fact). But maybe that debate will soon reach a conclusion as well. Note that Pearson seems to be going the async route in their new PEE Test, and that ETS will offer an async option in the new TOEIC Link Test (which is being pitched to higher-ed as an admissions test). These developments suggest that the writing is on the wall for live proctors. Indeed, I was a little surprised to learn that ETS will maintain them as part of the revised TOEFL set to launch in early 2026.

Check out this wonderful new article in TESOL Quarterly by Yasin Karatay and Jing Xu.  It explores the possibility of using an AI-powered spoken dialog system to simulate an IELTS examiner.  Specifically, the researchers used a self-developed SDS powered by GPT-4o to simulate the third part of the IELTS speaking test.

It is important to note that in this research the AI served only as the interlocutor.  Ratings were carried out by trained humans who reviewed recordings of the interactions.

The authors concluded that “the SDS consistently elicited some key interactional competence features, as seen in face-to-face oral proficiency interviews, and that such features were useful in distinguishing between higher- and lower-proficiency test takers.”  They also highlighted areas for improvement around non-verbal clues and some unnatural interactions.

There is a lot more to it, of course.  So take a moment to read the article.

Though the IELTS partnership takes a “cautiously curious” approach to the use of AI in testing, this may be an area worth exploring in more depth.  Some observers have raised concerns about the ability of the partnership to maintain consistent standards across four million annual speaking tests, each carried out by an individual examiner.  It is possible that at least some of these administrations are impacted by things like bias, fatigue and ability in the examiner population.  It’s no secret that IELTS test takers frequently travel to test centers which they feel are more conducive to a higher speaking score. The IELTS partnership is quick to dissuade test takers of this notion, but it isn’t inconceivable that one examiner might be better at their job than another.  Or that one might possess certain biases which another does not and that such differences could impact how the speaking test unfolds.

On top of that, this sort of change could lead to immense cost savings for the organizations that administer the IELTS (even if human raters are retained).  Some of those savings could be passed on to test takers, if only to make the IELTS more competitive in an increasingly crowded market of tests.  That’s a touchy subject so I’ll leave it for another day, but I think most readers can imagine the possibilities.

Finally, it is worth noting that this research was done by researchers at Cambridge University Press and Assessment.  So maybe this could lead to something.

You may also appreciate this similar research funded by ETS back in 2021.  Based on that work, I’m still optimistic that an AI-powered SDS system will find its way into some future version of the TOEFL.