I'd love to think I was wrong, JD -- and I probably am. It's based on the idea that AB tends to overestimate right-wing support in the country by about 20-25% (ie final result = AB prediction/1.25 or thereabouts), which would then swing (mostly) back to Labour.
In fact even this correction factor appears wildly optimistic, and is based on the EU referendum. In previous elections the AB predictions are usually within 3 points or so of the final outcome (at least for the Tories; UKIP support on AB is wildly out from national levels). So on that basis, if this is a continuation of the same pattern then AB has got Tory support too high (surely!), but still might indicate that the party can pick up not that far from 50% of the national vote, and mostly at the expense of Labour.
It's wildly unscientific to be sure, because I can't say I've bothered to correct for, for example, the number of people who voted in each AB poll -- and maybe some left-wing AB members haven't been as active lately, skewing the numbers a bit. But I'm putting this out there, for what it's worth: AB predicts a Tory landslide that puts the results of 1997 to shame, even if you attempt to correct for the right-wing bias of most AB polls.