About Influ

What Influ is, what it measures, and the one thing it does that a chat cannot: report how many of the sources that could have answered actually did.

What Influ is

Influ measures whether a political narrative is actually travelling.

You describe a claim or a campaign in plain language. Influ works out what to search for, goes and looks across twenty sources, counts what it finds, and scores it out of 100. Then it tells you how many of those sources actually answered, so you know how much the number is worth.

That last part is the whole point.

What it does, step by step

You write the narrative. One or two sentences in your own words, in any language. Or paste a link, or a picture of a poster.
An AI works out what to search for. Search terms, slogans, which side the campaign is on, and the words the other side is using. All of it is shown to you, and all of it is editable — if it has aimed badly, fix it and run again.
Influ goes and looks. YouTube, Google searches, Bluesky, Mastodon, Wikipedia, news, Telegram, ad libraries, official records and betting markets.
It scores what it found. Attention, whether people are positive or negative, and whether your own words are being repeated back.
It keeps the result. Every measurement is stored, so the next one can be compared against it. Set a cadence and Influ re-measures on its own, and tells you when something moves.

Why it is different

There is no shortage of tools that will tell you something about a campaign. Almost all of them are confident. That is the problem.

Influ tells you what it missed. Every result carries the number of sources that answered — "8 of 17". A run measured across a third of the sources is a weaker reading than one measured across all of them, and you are told which one you are holding. Nothing else here does that, and no chat model can: an AI answer reads exactly as confident whether it found everything or nothing.

Every number traces back to something. Real posts, real links, real counts from named sources. Not a model's impression of what people are probably saying.

It is repeatable. Same terms, same window, same sources, saved with a reference code. Run it again in a month and the difference is real. Ask a chat model twice and you get two essays that cannot be subtracted from each other.

It is a series, not a verdict. One score tells you how a narrative is doing. Ten scores tell you whether it is moving — which is the question people actually have, and the only one worth setting an alarm on.

The scoring is published. Attention 30%, sentiment 35%, message cut-through 35%. You can disagree with the weights. That is deliberate: a score you cannot argue with is not evidence.

What it is not

It is not a chatbot, and it will lose to one on an open-ended question. If you want a campaign explained to you, ask Claude or ChatGPT — they can follow a story anywhere, and Influ cannot.

It is not a firehose. It does not crawl everything and rank it. It measures a few things you have chosen, carefully, over time.

It will not tell you how to make a narrative spread. It measures what is happening, not how to manipulate it.

Where it is weak today

Sources outside the United States. Several of the sources cannot cover a non-US campaign at all — American government records have nothing to say about an Israeli election — but they are still counted against the coverage figure, which makes a perfectly good run look uncertain. One source returns nothing at all for Hebrew searches while working normally in English.

Both are known, and both are being fixed. In the meantime the coverage number is the worst case rather than the true one, and it is written on every result rather than hidden.

Open Influ