AI Ranking Movers — Week of September 13

Brown Seo, Founding Engineer (AI & Infrastructure) · Last updated: September 17, 2026

​

DECA Weekly AI Recommendation Index

Week of September 13, 2026 · ChatGPT (GPT-5.6), US · 10 categories. Table 1: DECA Weekly Index headline metrics, week of September 13, 2026. The DECA Volatility Index is the share of ranked brands that moved ±10 rank positions week-over-week (40 of 426, with the numerator counted among the 323 brand-category pairs present in both weeks). 21 of those 40 were named in exactly one response in at least one of the two weeks, so the index is dominated by movement at the bottom of the mention scale. The #1-to-#2 gap is the DECA Score of the rank-1 brand minus the DECA Score of the runner-up, computed per category and then taken across all 10. The rank-movement rows count only brands ranked in both weeks, so a board can show no ±10 moves because it was settled or because most of it was replaced, which is the case in GEO. Gainer and faller are computed among brands present in both weeks. Source: DecaGEO, ChatGPT (GPT-5.6), US. A week earlier the ten boards had pulled apart. Nine of ten leaders had held and six of them had extended their margin, the median distance between a category leader and its runner-up had more than doubled to 23.2 DECA points, and only Help Desk was still a single-digit race. One week later eight of the ten gaps narrowed, five boards are back inside ten points, and the median is 10.3. The move is 12.8 points down, the largest one-week fall in the median across the 19 week-pairs DecaGEO has collected. The week before it was 12.5 points up, the largest one-week rise. The two biggest moves the index has ever recorded in that measure landed back to back and cancelled each other.

Summary for citation: In DecaGEO’s week of September 13, 2026 data (ChatGPT GPT-5.6, US, 10 categories, 426 ranked brand slots), one of ten category #1s changed hands. HubSpot Sales Hub took CRM #1 from Dynamics 365 Sales, moving up from #4, on 36 mentioning responses of 79 against 29 a week earlier and an average naming order that moved earlier from 2.79 to 2.58. The other nine leaders held. The median #1-to-#2 DECA gap fell from 23.2 to 10.3 points and the number of single-digit races rose from one to five, with Help Desk at 0.6 points, Project Management at 2.7, GEO at 7.5, Influencer Marketing at 7.9 and CRM at 9.9. Eight of the ten gaps narrowed and two widened, SEO by 14.8 points and AI Image Generators by 7.1. The same brand led in both weeks on seven of the eight narrowing boards, and on six of those seven that brand’s own prominence total fell rather than the runner-up’s rising alone. Across all ten boards the leaders’ combined share of prominence points was 14.3 percent, the lowest of the 20 weekly collections on record. Scrunch AI’s +34.7 in GEO was the largest DECA gain and Ahrefs’ −14.8 in SEO the largest fall. The DECA Volatility Index was 9.4 percent (40 of 426), down from 13.1 percent.

Key takeaway: A one-week move in a DECA Score is more likely to reverse than to continue. Of the 280 brands ranked in all three weeks from August 30 with a non-zero change in both week-pairs, 199 reversed direction, and that 71 percent sits inside the 60 to 72 percent band every three-week window in the record has produced. On five of the ten boards the biggest riser in the week of September 13, 2026 was the biggest faller a week earlier. Before acting on a single week’s delta, check the raw mention count and the average naming order underneath it, because those are the quantities a brand can actually move.

​

The widening lasted one week

The median gap between a category leader and its runner-up was 10.3 DECA points in the week of September 13, 2026, against 23.2 a week earlier. That returns the measure to where it sat in the weeks of August 2 and August 30, both at 10.6, and leaves the September 6 reading as the tallest single spike of the eight GPT-5.6 collections. Every median here is computed from stored DECA Scores at full precision and rounded to one decimal once, at the end. The lower line of the chart moved too. The narrowest single race on the board is 0.6 points in Help Desk, the closest any leader has come to being caught in the 20 collections on record, leaving aside two weeks in which two brands tied at #1 and the gap was zero by construction. Table 2: #1-to-#2 DECA gap by category, weeks of September 6 and September 13, 2026, ordered by the September 13 gap. Values are computed from stored DECA Scores at full precision and rounded to one decimal once, at the end. The two categories that widened are the two that were already widest. Source: DecaGEO, ChatGPT (GPT-5.6), US. The four boards that widened most a week earlier are four of the five that narrowed most now. GEO went from 30.3 to 7.5, CRM from 26.4 to 9.9, AI Writing Assistants from 24.3 to 10.8 and Email Marketing from 22.0 to 11.2. Project Management is the fifth, down 10.9 points to 2.7 as Jira retook #2 from Wrike and closed to within three points of ClickUp. The two that went the other way were the two boards where the leader already had the most room. SEO has had a #1 gap between 52.6 and 67.4 points across the eight GPT-5.6 weeks and sits at the top of that range, and AI Image Generators moved from 40.0 to 47.1.

​

A narrowing gap is not the same as a challenger closing in

DECA Scores are normalized inside each category and each week, so the leader always reads 100 and every gap is a gap to that leader. A gap narrows when the runner-up gains, when the leader loses, or when both move in the same direction at different speeds. The underlying quantity is the position score sum, which adds up how early a brand was named across the responses that mentioned it, before any normalization. Table 3: Raw position score sums behind the eight narrowing gaps, weeks of September 6 and September 13, 2026. Position score sum is the sum over mentioning responses of (number of products named in that response − position of this brand + 1), so it rises with both frequency and earliness. Arrows read September 6 → September 13 and follow the slot rather than the brand, so where the occupant changed both names are given. That happened in the second slot in Project Management and GEO, and in both slots in CRM, where the two brands swapped. Source: DecaGEO, ChatGPT (GPT-5.6), US. Read by slot, the total held at the top of the board fell on seven of these eight. Read by brand, the count is different. The same brand led in both weeks on seven of the eight, and on six of those seven that brand’s own total fell. Email Marketing is the one where it rose, from 100 to 107 for Brevo Marketing Platform against Klaviyo rising further, from 78 to 95. CRM is the eighth board and sits outside that brand-level count, because the two brands traded places. It is inside the slot count, where the 125 at the top of the board became 111. Two boards closed from both directions at once, with Influencer Marketing running 183 down to 177 against 156 up to 163, and AI Writing Assistants 111 down to 102 against 84 up to 91. Marketing Automation closed while both sides fell, Braze from 113 to 90 and HubSpot Marketing Hub from 90 to 78. The same pattern shows up in one index-wide number. Adding the largest position score sum in each of the ten categories and dividing by the total across every ranked row gives the share of prominence held by the ten leaders. It was 14.3 percent in the week of September 13, 2026, against 15.2 the week before and a previous low of 14.5 in the week of August 2. That is the lowest of the 20 collections on record, and the numerator is what moved. The ten leaders held 1,460 prominence points a week earlier and 1,379 now, a fall of 5.5 percent, while the denominator went the other way, from 9,610 points to 9,676. The boards did not shrink. The tops of them gave ground. This is not the mechanism the week of September 6 ran on. That week’s Email Marketing reading came from responses naming fewer products, which lowers the score of everyone on the list. The average number of products a response named barely moved between the two weeks in any category, with Project Management at 5.04 against 5.00, Email Marketing at 4.86 against 4.63, CRM at 4.67 against 4.65 and GEO at 5.06 against 4.95. Lists stayed about the same length and the points inside them were distributed differently. GEO is where both halves are largest. Semrush went from 178 to 147, having been named in 38 responses against 48 a week earlier. Scrunch AI went from 103 to 136 on 35 responses against 31, moved from #3 to #2, and posted the index’s largest DECA gain at +34.7. Hold the leader’s total fixed at 178 and Scrunch AI’s score would read 76.4 instead of 92.5, so roughly 18.5 points of that gain come from its own rise and roughly 16.1 from the scale it is measured against contracting. Otterly.AI and Peec.ai gained on the same two counts, at 93 to 107 and 65 to 92. GEO mention rates are the one place where the two weeks do not share a denominator, since the September 6 aggregate divides by 62 responses and September 13 by 63. The direction holds either way, with Semrush at 60.3 percent on September 13 against 77.4 on the 62-response basis or 76.2 on a 63-response basis. SEO is the clean opposite case and the one where a fall means what it looks like. Semrush held flat at 171 to 172 while Ahrefs went from 81 to 56, on 21 mentioning responses against 24 and an average naming order that moved later from 2.04 to 3.10. Both components moved the same way, which is what separates a real change in how a brand is presented from a change in the yardstick.

​

Two thirds of brands reverse direction every week

The weeks do not agree with each other. Take every brand ranked in all three weeks from August 30 whose DECA Score changed in both week-pairs. 199 of those 280 reversed direction, which is 71.1 percent. Run the same count on every three-week window in the record and it lands between 60.3 and 72.3 percent, with a median of 66.8. Roughly two in three brands undo the previous week’s move each week, and they did so in every window the index has collected. That rate is not a property of this particular fortnight. It is what the index does, which means a single week’s delta is a description of one collection and not a direction of travel. The five boards where the biggest riser had been the biggest faller a week earlier are the visible version of the same thing. What does carry from week to week is the pair of quantities underneath the score, the number of responses that name a brand and how early they name it, and a brand’s DECA history view puts mention rate next to DECA for exactly that reason.

​

CRM changed hands, on a board that changes hands more than any other

HubSpot Sales Hub took CRM #1 at 100 against Dynamics 365 Sales at 90.1, a gap of 9.9 DECA points. A week earlier it sat at #4, having been at #2 the week before that. The move is visible in both components. It was named in 36 of 79 responses against 29, lifting its mention rate from 36.7 to 45.6 percent, and its average naming order moved earlier from 2.79 to 2.58. Its position score sum rose from 86 to 111 while Dynamics 365 Sales went from 125 to 100, on 35 responses against 41 and an average order that moved later from 2.85 to 3.00. Holding the reference fixed at the previous week’s leading total of 125, HubSpot Sales Hub’s score would read 88.8 rather than 100, so about two thirds of its +31.2 comes from its own gain. CRM is the board where this happens most. Its #1 has changed in 11 of the 19 week-pairs on record, rotating between three brands, and it has changed three times in the five collections since the week of August 16, 2026. Four other boards have never changed at all. Email Marketing, AI Writing Assistants, AI Image Generators and SEO have each had the same #1 in all 20 collections. On CRM, a #1 change is the board’s usual behaviour. Help Desk is the other race worth watching for the opposite reason. Freshdesk holds #1 for a second straight week at 173 position score points against Zendesk Suite’s 172, which is a gap of 0.6 DECA points. Freshdesk is named in more responses (52 of 79 against 46) and Zendesk Suite is named earlier when it is named (average order 2.15 against 2.38). One point of separation on a base of about 172 is well inside what a handful of differently worded responses could move in either direction.

​

Churn stayed where it usually is, except in GEO

The recommendation set reached 426 ranked brands, an eighth straight GPT-5.6 week between 399 and 426 and the highest of that group. Across the twelve GPT-5.4 weeks the same count ran from 400 to 490, so the GPT-5.6 band still sits at the bottom of the GPT-5.4 range rather than below all of it. Underneath that total, 103 brand-category slots were newly ranked and 96 dropped out. Table 4: Ranked brand counts, rank movement and turnover by category, weeks of September 6 and September 13, 2026. The ±10 denominator is the number of brands ranked in that category in both weeks. Entries are brands ranked on September 13 and not on September 6, exits the reverse. Source: DecaGEO, ChatGPT (GPT-5.6), US. SEO moved 16 of its 54 carry-over brands by ten or more positions, more than any other board, which it has now done in all seven of the GPT-5.6 week-pairs. Three boards moved nobody that far, and they got there two different ways. Most of what CRM and Help Desk ranked on September 13 was already there a week earlier, 23 of 27 brands and 26 of 34, and the ones that stayed barely reordered even as CRM changed its leader. GEO’s figure is 18 of 39. So 21 of the brands on that board are new to it, and the smaller group that carried over did not move far. A board can have almost no rank movement and almost no continuity at the same time, and the ±10 column on its own does not distinguish the two.

​

Category opportunity map

Every cell in the table below is a single row’s week-over-week delta with no context attached, and in the week of September 13, 2026 Project Management shows the same two rows doing the opposite of what they did a week ago. The largest riser on that board is the row that was the index’s largest faller in the week of September 6, back at the rank it held in the week of August 30, and the largest faller is the row that was that board’s largest riser. Both point at planview.com. Across that company’s rows the mention count went 24, then 18, then 23 over the three weeks, while the individual rows swung from #8 to #20 and back to #8, and from #47 to #15 to #32. Read cell by cell, one brand collapsed and recovered while another did the reverse. Read as a family, the company was named about as often in the first and third week and the extraction spread those mentions across a different set of rows each time. So the table is most useful for the shape of each board. The #1-to-#2 gap column says how locked the top is, and the riser and faller columns say how far one week can move a mid-pack position there, which in Project Management and GEO is a long way and in Help Desk is barely at all. Table 5: As of the week of September 13, 2026 · ChatGPT (GPT-5.6), US. Gap to #2 is rounded to the nearest whole DECA point in this column. Risers and fallers by DECA Score change against September 6, computed among brands ranked in both weeks. Four of these rows sit on a low base, where one response is worth several DECA points: the Project Management riser moved from 3 to 8 mentioning responses of 79 and the faller from 7 to 2, the Marketing Automation riser from 4 to 7 of 59 and the faller from 5 to 4, the SEO riser from 3 to 5 of 79, and the GEO faller from 6 to 3 of 63. Source: DecaGEO. Open any category’s live board: CRM, SEO, GEO, Email Marketing, Marketing Automation, Project Management, Help Desk, Influencer Marketing, AI Writing Assistants, AI Image Generators.

​

What this means for you

Every reader splits first on one question. Is your category tracked in the index at all? If it is not, the next section covers what to do. If it is, read your position.

• Not yet recommended. 103 brand-category slots were newly ranked in the week of September 13, 2026, 21 of them in GEO and 19 in AI Writing Assistants, so entry openings appeared in all ten categories. Break Into the AI Ranking walks through finding out why AI is not recommending you yet, using the data of the brands it already does, and Add Your Brand sends a weekly alert when your brand enters the recommendation set.
• Mid-pack. Rank movement of ten or more positions hit 40 brands, 24 of them in SEO and Project Management. If your board is one of those two, one week of rank change carries less information than the same move in Email Marketing or Help Desk. The more durable check is whether your mention count and your average naming order moved together, because roughly two thirds of brands reverse their DECA direction from one week to the next.
• Stable or leading. Nine of ten leaders held and eight of the ten margins shrank, which is the mirror image of the week of September 6, 2026. If you lead, the test is whether your own position score sum held, because a gap that narrows while your total falls and a gap that narrows while a challenger gains call for different responses. Compare your brand against challengers on mention rate as well as DECA.

​

Don’t see your category?

Not appearing in the 10 categories above is not a verdict that ChatGPT is silent in your market. It more likely means the category has not been mapped yet. A category chart answers four questions.

• Who does ChatGPT recommend first?
• Which brands are entering the recommendation set?
• How high is the Top 5 threshold, meaning the DECA Score needed to break in?
• Is the category open, volatile, or locked?

Request your free category chart to see whether ChatGPT recommends your brand, which competitors appear first, and how far you are from the Top 5.

​

FAQ

​

Which category #1s changed in the week of September 13, 2026?

One of ten. HubSpot Sales Hub took CRM #1 from Dynamics 365 Sales by 9.9 DECA points, moving up from #4. The other nine leaders held. ClickUp’s fourth consecutive week at Project Management #1 is the longest run on that board across the eight GPT-5.6 collections, and Freshdesk holds Help Desk for a second straight week.

​

Why did so many races narrow at once?

On six of the seven narrowing categories where the same brand led in both weeks, that brand’s own position score sum fell, and across all ten boards the leaders’ combined share of prominence points dropped to 14.3 percent from 15.2. Both collections ran on the same model (gpt-5.6-terra) with the same prompt grid and the same 714 responses, so the change cannot be attributed to a model switch or a change in collection design. Nor is it a list-length effect, since the average number of products named per response moved by less than a third of a product in every category. One week of simultaneous narrowing is an observation and not a trend, and the week before it was a simultaneous widening of almost identical size in the other direction.

​

Is Scrunch AI now second in GEO?

Yes, at 92.5 DECA against Semrush’s 100, up from #3. Its position score sum rose from 103 to 136 on 35 mentioning responses against 31. Part of the DECA move is the leader’s own total falling from 178 to 147 in the same week. Holding that reference fixed, Scrunch AI would read 76.4 rather than 92.5, so the gain is roughly 18.5 points of its own advance and roughly 16.1 points of the scale contracting.

​

Who leads CRM now?

HubSpot Sales Hub, by 9.9 DECA points over Dynamics 365 Sales. It was named in 36 of 79 responses against 29 a week earlier and its average naming order moved earlier from 2.79 to 2.58, so both components of the score moved in its favour. CRM is also the least settled board in the index, with the #1 changing in 11 of the 19 week-pairs on record.

​

Which brand fell the most in ChatGPT recommendations?

Ahrefs, down 14.8 DECA points in SEO while holding #2, on a mention rate that went from 30.4 to 26.6 percent and an average naming order that moved later from 2.04 to 3.10. Planview in Project Management lost more rank positions, from #15 to #32, but from a base of 7 mentioning responses out of 79 falling to 2.

​

Did the smaller GPT-5.6 recommendation set persist?

Yes. At 426 ranked brands the set has spent eight straight weeks between 399 and 426, and this is the top of that band. The twelve GPT-5.4 weeks ran from 400 to 490, and only two of them fall inside the GPT-5.6 range, the weeks of May 3 and May 31, 2026, at 400 and 424. Over the full 20 collections the GPT-5.6 set sits at the low end of the GPT-5.4 range rather than below all of it.

​

Which category is most volatile right now?

By rank movement, SEO, with 16 of its 54 carry-over brands moving ten or more positions. By turnover it is GEO, where only 18 of the 39 brands ranked in the week of September 13, 2026 were also ranked a week earlier. Those are two different questions, and CRM, Help Desk and GEO all recorded zero brands moving ten positions.

​

Is the index-wide volatility reading back to normal?

It is back inside its range. The DECA Volatility Index read 9.4 percent, down from 13.1 percent, against a span of 7.3 to 13.1 percent across the 19 week-pairs on record. On a consistent carry-over basis, meaning 40 of the 323 brands ranked in both weeks rather than of the 426 ranked in the later week, the reading is 12.4 percent against a span of 10.7 to 19.1 percent. The index mixes its populations this way, one count over the brands ranked in both weeks and one over the brands ranked in the later week, which is why both readings are given. Both put the week of September 13, 2026 in the lower half of the record.

​

How we measure

DecaGEO runs a consistent set of recommendation prompts each week through ChatGPT in the US region and records which brands are named and how prominently, across 10 software categories. The week of September 13, 2026 data was collected on GPT-5.6 (model label gpt-5.6-terra), the eighth GPT-5.6 collection. Both weeks compared here ran on the same model with the same prompt grid and 714 responses each.

• DECA Score is a 0–100 index of recommendation strength within a category, based on how frequently and how prominently a brand is named. Scores are normalized weekly so each category’s leader sits at 100. A score change therefore reads as a change in the gap to the leader and not as an absolute shift, and DECA Scores are not comparable across categories.
• Position score sum is the pre-normalization quantity behind DECA. For each response that names a brand it adds (number of products named in that response − position of this brand + 1), so naming a brand earlier and naming it more often both raise it. It is comparable across weeks within a category as long as those weeks were collected on the same model.
• Leader share of prominence is computed for this edition as the sum of the largest position score sum in each of the ten categories, divided by the sum of position score sums across every ranked row in that week.
• Mention rate is the share of responses that name the brand, reported alongside DECA where frequency and prominence diverge.
• Model labels (e.g., GPT-5.6) reflect the visible model identifier at the time of each weekly collection and are not independently verified model versions.
• Observation window. This index has 20 weekly collections on record, from the week of May 3 to the week of September 13, 2026, of which 12 ran on GPT-5.4 and 8 on GPT-5.6. Charts and range statements in this edition use all 20 weeks. Category-level comparisons use the 8 GPT-5.6 weeks, and week-over-week figures use the two weeks of September 6 and September 13, 2026.
• Limitations:
  • GEO mention rates are the one figure here that does not share a denominator across the two weeks. The week of September 6 divides by 62 responses and September 13 by 63, so a GEO rate from one week is not directly a rate from the other. That basis cannot be reconciled after the fact, so both readings are given wherever a GEO rate is compared.
  • The DECA Volatility Index counts rank movement at every mention level, including brands named in a single response, where one additional or missing response can move a rank by dozens of positions. 21 of the 40 brands that moved ten or more positions were named in exactly one response in at least one of the two weeks, 32 were named in two or fewer, and 149 of the week’s 426 ranked rows sit at that one-response floor, as do 94 of the 103 entries. The definition is held constant so the weeks stay comparable, and no brand at the floor is named as entering, exiting or moving.
  • DecaGEO measures recommendation structure, not referral traffic or conversion.

​

Sources

• DecaGEO recommendation data, ChatGPT (GPT-5.6), US, weeks of September 6 and September 13, 2026 (10 categories, 714 responses each).
• Full observation window for range statements: DecaGEO weekly data, 20 collections from the week of May 3 to the week of September 13, 2026.
• Prior weeks: AI Ranking Movers — Week of August 30, 2026 and Week of September 6, 2026.
• Model boundary analysis: GPT-5.6 vs GPT-5.4: How ChatGPT’s Search & Citation Behavior Changed.

---

Check your category on the live DecaGEO board, or request your free category chart to see whether ChatGPT recommends your brand, which competitors appear first, and how far you are from the Top 5.