SEO Performance Dashboard

7,521 users, down 12.7% on July. The fall is concentrated in direct traffic, 4,932 down to 4,133. Organic search eased to 1,733, down 6.9% on July but still 14.8% ahead of June's 1,510. Paid search continued its decline to 1,475. AI assistants sent 56 users, up from 19 in July, tripling month on month.

Organic produced 144 key events at a 5.21% session rate, against 89 from paid at 4.01%, on 1,733 organic users to paid's 1,475. Across the quarter organic key events have held roughly flat, 158 to 141 to 144, while paid has fallen from 162 to 89. Almost all organic key events are technical downloads, and they are specification-stage documents rather than general brochures: the Forte LEC product data sheet, the ConCore 1500 and 1250 panel data sheets and the Forte LEC installation manual were the most downloaded by organic visitors in August. Enquiry-form events remain too few to draw a trend from, at one lead and one form submission. Two known gaps sit behind that: reCAPTCHA breaks part of GA4's automatic tracking, and the Google Ads conversion tag fix built on 22 July is staged in Tate's GTM container as three unpublished changes awaiting Tate's publish.

1,118 clicks, down 9.3% on July, against 52,192 impressions, level with July. CTR fell to 2.14% and average position to 14.1. The CTR fall is mostly a change in the mix of queries Tate appears for. Named non-branded impressions rose again to 17,188, double June's 8,622, and non-branded search converts at 0.24% against 4.65% on branded terms. Non-branded went from 19.3% of impressions in June to 32.9% in August. Holding June's click-through rates and changing only the mix accounts for 0.19 of the 0.32 point fall in CTR; holding June's mix and changing only the rates accounts for 0.05. Roughly two thirds of the decline is Tate appearing for more non-branded queries, not Tate performing worse on the queries it already had. The same mechanism moved average position: non-branded terms sit at a weighted average of 17.7 while branded terms sit in the low single digits, so a heavier non-branded mix drags the site-wide average down without any individual page losing ground. The concentration is stark. Liquid cooling returned 1,143 impressions and 7 clicks at position 6.1, concore returned 1,023 impressions and 1 click at position 6.1, and liquid cooling data center returned 408 impressions and 2 clicks at position 8.9. Those three terms produced 2,574 impressions and 10 clicks between them.

35 tracked, 17 on page one, 2 climbing, 14 in AI Overviews. Data centre cooling entered page one for the first time, moving from unranked to position 3. Three high-volume terms dipped mid-month: raised flooring, at 590 searches a month, and liquid cooling data centre, at 110, fell out of the rankings, and liquid cooling system eased from position 1 to 5. All three had returned to position one at the 2 September check. Three terms have since dropped out of the rankings between the August and September checks, including data centre cooling itself: data centre cooling from 3 to unranked, data centre cooling systems from 3 to unranked, and data centre liquid cooling from 1 to unranked. Those are the September watch list. On cause, the data does not support a single answer. The August spam update ran inside the window, but the worst position days were 26, 29 and 30 August, after it had finished, so it is a possible contributor rather than a confirmed one.

14 of the 35 tracked terms carry a Tate citation in AI Overviews, 40% of the set. A live check on 2 September found Tate cited in 5 of 6 broader Australian commercial searches; the miss was "raised access floor supplier sydney", where a third-party distributor was cited instead. AI assistants sent 56 users to the APAC pages in August, up from 19 in July, with ChatGPT accounting for 48. AI referral traffic is growing faster than any other channel on the site. One limit worth stating: the keyword tracker records a single AI Overview status per term, not a month-by-month history, so month-on-month change in AI Overview presence cannot be evidenced from it.

49% of tracked terms on page one, 6% climbing, 40% cited in AI Overviews.