The National Planning Policy Framework, August 2026
On 17 August 2026 the government published the biggest rewrite of national planning policy since 2012, and Annex A made it material for decision-taking from the day of publication. No transitional period, no grace window. Here are the headlines every agent needs this week - and below them, the less obvious points that will actually decide cases.
In force from 17 August 2026
That last point is the one to act on. A planning statement filed in July citing paragraph 11(d) will be determined by an officer reading policy S4 or S5. The argument usually survives translation; the citations do not. If you have live applications, a short addendum re-basing the case on the new policy codes is the cheapest insurance you can buy this month.
The presumption in favour of sustainable development
Under the December 2024 Framework, one question dominated: is the plan up to date, and can the council show a five year housing land supply? The August 2026 edition asks a different first question: is the site inside a settlement? Everything else follows from the answer.
Approve unless the benefits are substantially outweighed by adverse effects.
Supply, plan age and the HDT switch nothing on here.
Only the closed list at S5(1)(a) to (j) is supported.
Housing needs a gateway: the supply triggers live at S5(1)(j).
Inappropriate development needs very special circumstances (GB6). The grey belt route is GB7(1)(g) - an all-of test with its own supply trigger at limb (ii), via footnote 41.
Not-inappropriate development keeps the presumption - S5(5).
On both routes, the protective policies are folded into the "substantially outweighed" test rather than standing in front of it as a separate gateway. S4(2)(a)(ii) names Local Green Space, Protected Landscapes, biodiversity and the rest; S4(2)(c) and S5(2) catch failure to comply with any national decision-making policy that requires refusal in specific circumstances. The old asset list does survive, at footnote 24 - but it is anchored at S1(1)(a)(i) and scoped to plan-making, so it is not a decision-side gateway. In practice the Monkhill sequence is unchanged: run the protective test first, because failing it is what makes the benefits substantially outweighed.
The practical consequence, and the sentence worth sharing: inside a settlement, your case no longer turns on whether the council can demonstrate a five year housing land supply. S4 applies permanently and unconditionally. A collapsing HDT unlocks nothing extra on an urban infill site - and equally, a healthy supply position no longer protects the council from it.
We mapped what that looks like across England, district by district and settlement by settlement: the presumption map, before and after.
From paragraphs to policy codes
Every paragraph number you know is gone. The Framework is now 131 coded policies across 20 chapters, plus six annexes that are themselves policy. Two chapters are purely procedural (PM for plan-making, DM for decision-making); the thematic chapters mix plan-making and decision-making policies, and the split matters - more on that below.
Plan-making chapter. Within the thematic chapters, each policy is individually labelled plan-making or decision-making - citing the wrong half is wrong law, not a style slip.
Four concepts removed from the Framework
December 2024 to August 2026: the key citations
Where the substance moved unchanged, translate and move on. Where the row is marked in red, the test itself changed - read the new policy before relying on the old argument.
| December 2024 | August 2026 | Notes |
|---|---|---|
| para 11(d) | S4 / S5 | Changed, not renamed: the presumption routes by location. Inside a settlement, supply is irrelevant to the test; outside, the triggers live at S5(1)(j) |
| footnote 7 | S4(2) / S5(2) / S5(5) | Changed, and easy to get wrong: the gateway that disapplied the presumption is gone, and its work now sits inside the "substantially outweighed" test. The asset list itself survives at footnote 24, but anchored at S1(1)(a)(i) and scoped to plan-making |
| footnote 8 | footnote 27 | The 5YHLS and HDT triggers, now anchored to S5(1)(j) |
| para 14 | S6 | Neighbourhood plan protection, reframed in the S4/S5 grammar |
| paras 48-49 | DM4 | Emerging plan weight at DM4(1)(a)-(c); prematurity at DM4(2)-(3) |
| para 61 | Annex D 1-7 | Standard method, formula unchanged |
| para 78 | Annex D 8-9 | Five year supply and buffers, unchanged |
| para 79 | Annex D 12 | HDT bands identical (95/85/75), but the sub-75% consequence now feeds S5(1)(j) as deemed unmet need - outside settlements only |
| para 80 | Annex D 13 | New results apply the day after publication - the 2025 measurement is already operative |
| para 155 | GB7(1)(g) + Annex E | Grey belt gains a dedicated assessment methodology, and the definition changed (see below) |
| paras 156-157 | GB8 | The Golden Rules keep their name |
| para 212 | HE6(1) | "Great weight" is now "substantial weight", irrespective of harm level |
| para 215 | HE6(3)-(4) | The balance survives; the two-tier vocabulary does not |
| para 216 | HE7(2) | Non-designated heritage assets: balanced judgement, unchanged |
Ten less obvious changes
The headlines above will be everywhere this week. What follows is the layer below: the points that decide real cases, in rough order of how much money each one moves. We re-based GapSense on the new Framework the day it published and verified every mapping against the published text - these are the points that tripped real assessments first.
Supply survived twice
The obvious reading of the new presumption - supply only matters outside settlements, via S5(1)(j) - misses the second gateway. On grey belt land in the Green Belt, GB7(1)(g)(ii) requires an evidenced unmet need, and footnote 41 defines that need using the identical two triggers: no five year supply (with the applicable buffer), or an HDT below 75%. So "it is Green Belt, therefore the supply position is irrelevant" is now wrong twice over. And everywhere - including inside settlements - a genuine shortfall remains a weighty benefit in the ordinary balance.
Grey belt is designation-blind
The new Annex B definition is short: grey belt is land in the Green Belt comprising previously developed land and/or any other land that does not strongly contribute to Green Belt purposes (a), (b) or (d). Three consequences agents are getting wrong already: previously developed status is not required (a greenfield paddock can be grey belt); a strong contribution to purpose (c) - safeguarding the countryside from encroachment - cannot disqualify land, because the definition never asks about it; and the old designations carve-out is deleted, so a listed building or SSSI no longer blocks grey belt status. It bites later, through its own policy, in the balance.
Do remember GB7(1)(g) is an all-of test: grey belt status is only limb (i), which also asks whether development would fundamentally undermine the remaining Green Belt. Unmet need (ii), a sustainable location with particular reference to TR3 (iii), and the GB8 Golden Rules for major housing (iv) all have to hold too.
S5(5) has a second sentence
Everyone will quote the first half: in the Green Belt (and on Local Green Space), S5 does not apply, and the case is decided under HC8, GB6, GB7 and GB8. Fewer will quote what follows: development which would not be inappropriate there gets the same approve-unless-substantially-outweighed presumption as everywhere else. A scheme that clears GB7 does not merely survive - it argues from strength.
Your local plan just got weaker
Annex A(2): a development plan policy which is materially inconsistent with the national decision-making policies now gets "very limited weight", unless the plan was examined against the new Framework - which today means almost no plan in England. The final softened the consultation draft's "in any way inconsistent" to "materially inconsistent", so this is a conflict-in-outcome test, not a vibes test.
Two counterweights before anyone declares the development plan dead: the same provision says a policy is not downweighted merely for predating the Framework; and settlement-boundary policies just became more load-bearing, not less - the whole S3/S4/S5 routing runs through the plan's settlement definitions.
75% exactly is not below 75%
The Housing Delivery Test consequences are unchanged in shape - below 95% means an action plan, below 85% adds the 20% buffer, below 75% now deems unmet need for S5(1)(j). The boundary matters: a council at exactly 75% takes the buffer but does not trigger the deemed need. With the 2025 measurement published the same afternoon, at least one authority sits precisely on that line - North Hertfordshire, at 75%.
Annex D 12(a)-(c). Each band runs up to, but not including, its upper figure - so 75.0% takes the buffer and 74.9% deems the need.
Half of it is not for decisions
Every thematic chapter mixes plan-making and decision-making policies, and citing the plan-making half in a decision is wrong law that reads as right law. Three pairs are already tripping people:
| Looks right, is wrong | The decision-making cite | Subject |
|---|---|---|
| N1 (plan-making: environmental opportunities) | N6(2) | Irreplaceable habitats - ancient woodland, ancient and veteran trees. Refuse unless wholly exceptional, irrespective of designation |
| TR2 (plan-making: local parking standards) | TR4, refusal test at TR6(4) | Street design, access and parking in decisions; the severe-impact and highway-safety refusal test |
| F1-F3 (plan-making: flood strategy) | F4-F8 | Site-specific flood risk, sequential and exception tests, sustainable drainage |
Heritage loses its two-tier vocabulary
Losing the two-tier vocabulary is not a relabelling. HE6(1) gives conservation substantial weight whatever the harm level; HE6(3) makes any harm to significance a matter of considerable importance and weight, balanced against public benefits at HE6(4). The graded tests survive only at the top end - substantial harm and total loss, HE6(5)-(6). If your heritage statement's architecture is "the harm is less than substantial, therefore...", it is arguing in a vocabulary the decision-maker no longer uses.
Data centres are now Essential Infrastructure
Quietly, in Annex F Table 2, the flood-risk vulnerability classification gained a whole new bullet of Essential Infrastructure: data centres, heat networks, hydrogen production, carbon capture and EV charging stations. Telecommunications installations moved out of Highly Vulnerable into the same row, and electricity generation and storage - already Essential under the old edition - are now spelled out explicitly. The classification decides which flood zones a use can occupy and how the sequential and exception tests run, so this is material for anyone promoting digital or energy infrastructure on lower-lying land. Given how battery storage has been faring at appeal, the table arriving in the same edition that rewrote the Green Belt tests is worth a careful read.
A village is not always a settlement
The Annex B settlement definition includes cities, towns, villages and predominantly built-up areas, plus allocated and permitted land forming part of them. It expressly excludes hamlets and scattered houses - and villages washed over by Green Belt. So the familiar argument "the site is within the village, therefore the in-settlement policies apply" fails precisely where it used to be deployed hardest. A washed-over village routes to the Green Belt chapter, not to S4.
The Framework translates itself
When the drafters carried the December 2024 grey belt policy forward, its cross-reference "with particular reference to paragraphs 110 and 115" became GB7(1)(g)(iii)'s "with particular reference to policy TR3". That is the model for handling an officer report or an appeal decision written under the old edition: translate the citation by substance, state the new code alongside the quoted old one, and never re-argue a point that carried over unchanged. Decision-makers will be doing exactly this for years of back-catalogue material - the agents who make their translations explicit will be the ones officers trust.
Sources and method
Everything above is taken from the published text of the National Planning Policy Framework, August 2026 edition (published 17 August 2026), read against the December 2024 edition it replaced. Policy codes were verified against the published PDF, not the December 2025 consultation draft, which renumbered several chapters before the final. The concordance is the same mapping we use in production: GapSense reads the current Framework directly, so its reports and this page cannot drift apart.
The usual caution. This is a briefing, not advice on a specific case. The Framework's transition provisions for plan-making are more forgiving than for decision-taking, several PPG topics still describe the old edition, and how inspectors handle the new tests will only become clear as decisions land. Where a live application turns on one of the points above, read the policy itself - it is short, and it is the law of the next decade.
Every GapSense report already speaks the new Framework.
We re-based our assessments on the August 2026 edition the day it published. A pre-submission check reads your draft application against the new policy codes, the local plan and the decision record - and flags the arguments that no longer exist. No card needed.