Most Monday mornings I do the same thing. Coffee, then I open my own RankingLocal weekly report for Yellow Pencil before I open anyone else's. Not because I expect drama — because the report is short, and I'd rather find a regression myself than hear about it from a client.
A lot of customers tell me they get nervous when the email lands. The score moved. A query dropped off. A competitor showed up where they used to be cited. I get it. When you've paid for a tool that measures something this specific, every wobble feels like a fire.
It usually isn't. This post is how I read the weekly report — mine and the ones I help clients read — without panicking, and without ignoring the things that actually matter.
What's actually in the email
The weekly email has six blocks, in this order:
- Your GEO Score this week, and the delta vs last week.
- A dimension breakdown — which of the four pillars moved, and by how much.
- Top 3 gaps (prompts where you should be cited but aren't).
- Top 3 wins (new or improved citations since last week).
- Competitor movement (who gained ground on your tracked prompts).
- Flare's note — a short plain-English read on what happened.
That's it. No dashboards to click through if you don't want to. The point of the email is that you can read it standing up with one hand on a coffee cup and know whether this is a fix-today week or a keep-shipping week.
Reading the score change
The GEO Score is a 0-100 number. It moves. Mine has swung 4 points in a week for no reason I could pin down, then swung back the next Monday.
Here's the rule I use: a swing of 3 points or less, in either direction, is usually crawler timing. The assistants we test against don't all refresh their indexes on the same schedule. If Perplexity re-crawled your site Sunday night and ChatGPT didn't, you'll see movement that has nothing to do with your content.
A swing of 5 points or more, sustained two weeks in a row, is a real signal. That's when I go digging.
Rule of thumb: don't chase the score. Chase the structural fixes. The score catches up in 2-3 weeks, not 2-3 days.
I've watched customers rewrite a page three times in a week trying to nudge the number. It doesn't work that way. The models cache. The crawl happens when it happens. If you've made a real fix, give it a fortnight before you decide it didn't work.
The dimension breakdown
We score four pillars: Authority, Clarity, Freshness, and Coverage. The weekly report tells you which moved.
If Clarity dropped, it's almost always a structural change on your site — you removed a heading, reshuffled an FAQ, a schema block broke on deploy. These are the easiest to fix. Twenty minutes with a developer.
If Authority dropped, something changed off-site. A citation on a directory went stale, a press mention got paywalled, a partner dropped your link. Slower to fix, but also slower to hurt you. I don't react to a one-week Authority dip.
Freshness is the sneaky one. If you haven't published or updated anything in six weeks, the models start de-prioritising you for queries that imply recency ("best … in 2026", "recent reviews of …"). The fix is boring: ship something. Doesn't need to be a novel. A 400-word update post with a real date on it moves this dimension more than people expect.
Coverage moves when a competitor starts getting cited for a prompt cluster you used to own. That's the one I always check first.
Top 3 gaps and top 3 wins
The gaps list is where I actually spend my time. These are prompts where, based on your topic map, you should be cited — and you aren't, or you used to be and you're not anymore.
For each gap, the report tells you which assistant isn't citing you, what it's citing instead, and a one-line Flare suggestion. Nine times out of ten the suggestion is "your page doesn't answer the specific sub-question being asked." Not that the page is bad. That it's answering a slightly different question.
The wins list is morale. Read it. Screenshot it. Send it to whoever on your team wrote the page that won. Then move on.
Competitor movement
This is the section that causes the most Monday-morning panic, and the one I'd most like customers to read slowly.
If a competitor gained on one prompt, that's noise. If they gained on three prompts in the same cluster, they shipped something. Go look at their site. Usually it's a new comparison page, a new case study, or they fixed their schema. You can see it in ten minutes.
Don't mirror them. Don't write the same page they wrote. Write the better version of the page you were already going to write, and ship it this week.
Flare's note
Flare is our advisor persona. The note at the bottom of the email is 80-120 words of plain English: here's what moved, here's what I'd do about it, here's what I'd ignore.
If you only read one block of the report, read this one. It's the part that translates the numbers into a decision.
A real Flare note from my own report two weeks ago read, roughly: "Score down 2, all in Freshness. No regression on tracked prompts. You haven't published since the 3rd. Ship the Phase 3 post you drafted and this reverses by next Monday." It did.
A sample snippet from Yellow Pencil's report
Here's an actual block from one of my recent weeks, cleaned up:
GEO Score: 71 (-2 vs last week) Authority: 78 (flat) Clarity: 82 (flat) Freshness: 58 (-6) Coverage: 66 (+1) Top gap: "web design agencies in Victoria BC" — Perplexity now citing 2 competitor directory pages. Your /about page is not being surfaced. Suggest: add service-area schema. Top win: "shopify plus agency canada" — ChatGPT now cites your Shopify Plus case study (new this week). Flare: Freshness dip is the full story. No structural issues. Publish the drafted post and re-check next Monday.
Two points down, one genuine to-do, one genuine win. Total reading time: under a minute. That's the shape of a normal week.
What to fix, what to ignore
Fix: a score drop of 5+ that's sustained two weeks. A specific tracked query that stopped citing you. A dimension drop tied to something you can name (a deploy, a content change, a link that broke). A competitor gaining on three+ prompts in the same cluster.
Ignore: single-week wobbles under 3 points. One assistant swapping your citation for a near-identical source. A gap on a prompt that isn't in your actual ICP. Anything Flare explicitly tells you is crawler timing.
The job is to stay boring. Ship structural fixes. Don't refresh-chase. Read the email on Monday, decide the one thing you'll do this week, close the tab.
If you want this report for your own business, pricing is here. If you've already got it and the Monday email still makes you anxious, forward me one — hello@rankinglocal.ai is read by me directly.